# IndustrialProcessAI.com > Free buyer's toolkit for industrial AI and machine learning: a weighted buyer's scorecard, an RFP question bank, a data-readiness assessment and a pilot-to-production gate review, plus plain-language guides. Editions in English (root), French (/fr/), German (/de/), Spanish (/es/), Italian (/it/), Dutch (/nl/), Portuguese (/pt/), Polish (/pl/), Czech (/cs/), Slovak (/sk/), Hungarian (/hu/), Romanian (/ro/), Swedish (/sv/), Turkish (/tr/), Chinese (/cn/). Coverage varies by language; every page lists its translations. This file holds the full text of every English page, followed by the index of translated pages. Figures are as published on the pages; sources are listed on each page. --- # Industrial AI Buyer's Toolkit: scorecard, RFP questions, readiness checks URL: https://industrialprocessai.com/ Summary: Free tools to evaluate industrial AI before you buy: a weighted buyer's scorecard, RFP question bank, data-readiness check and pilot gate review. Translations: fr https://industrialprocessai.com/fr/, de https://industrialprocessai.com/de/, es https://industrialprocessai.com/es/, it https://industrialprocessai.com/it/, nl https://industrialprocessai.com/nl/, pt https://industrialprocessai.com/pt/, pl https://industrialprocessai.com/pl/, cs https://industrialprocessai.com/cs/, sk https://industrialprocessai.com/sk/, hu https://industrialprocessai.com/hu/, ro https://industrialprocessai.com/ro/, sv https://industrialprocessai.com/sv/, tr https://industrialprocessai.com/tr/, zh https://industrialprocessai.com/cn/ The buyer's toolkit for factory AI ## Evaluate industrial AI on your line, not in a demo. Six weighted criteria, nineteen RFP questions, twenty-three data checks and a twelve-question gate review. Free Excel tools for plant, quality and CI leaders. Get the Buyer's Scorecard The six criteria → Start with measurementMost industrial-AI projects stall on data, not models. Measure first: which stops, how long, on which machine. TEEPTRAK does that with a module in the electrical cabinet and a sensor at the line. No PLC changes, live OEE in about two hours. The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → ### Free tools What an OEE system really costs: the 3-year calculator What an OEE system really costs over 3 years: licences, devices, sensors, screens, installation, integration and your team’s time. Free calculator. The Buyer's Scorecard: score your shortlist on six weighted criteria Free Excel scorecard: evaluate up to three industrial AI vendors on the six weighted criteria that predict deployment success. Nineteen questions vendors cannot answer with a brochure Grouped by the six rubric criteria we publish, each with what a substantive answer actually contains, so you can tell a real answer from a fluent… Score your data before you score the vendors Twenty-three checks across signal coverage, sampling, naming, history and governance: weighted, with a verdict at the bottom. Most failed pilots… Twelve questions between a working pilot and a line that depends on it Six gates (value, data, integration, people, commercial, risk) each with the evidence that settles it and the owner who has to produce it. Pass or… IATF 16949 §8.5.1.5 total productive maintenance: what auditors expect on OEE, MTBF and MTTR What IATF 16949 clause 8. 5. 1. 5 asks for on total productive maintenance, what evidence auditors look at for OEE, MTBF and MTTR, and where Excel… OEM capacity evidence: capacity is a standing, portal-reported obligation Mercedes SMB/BKM, IVECO clause 17, Volvo plant capacity assessment and Stellantis escalation: what evidence backs a capacity figure, and what fails. ### Start with a question What is an industrial AI platform? A plain-language definition of industrial AI platforms, the four platform categories, what they cost to deploy in time and people,… AI in manufacturing: 12 examples that actually produce ROI Concrete 2026 examples of AI and machine learning in manufacturing: scrap reduction, predictive maintenance, energy optimization,… Generative AI in manufacturing: separating the layer that works from the hype Where generative AI genuinely helps manufacturers in 2026, and where industrial machine learning, not LLMs, does the real process… The tools are free and work whatever platform you choose. To see where your lines lose time, talk to TEEPTRAK. Start here Score your data before you score the vendors Free, no login. One follow-up from TEEPTRAK, then nothing.Get the assessment → --- # AI in CEE manufacturing: why the data layer has to come first URL: https://industrialprocessai.com/cee/data-before-ai/ Summary: Only 3-8% of manufacturers in Romania, Poland and Hungary use AI (Eurostat 2025). Why industrial AI stalls on data, and a five-stage roadmap to fix it. Translations: pl https://industrialprocessai.com/pl/cee/najpierw-dane-potem-ai/, cs https://industrialprocessai.com/cs/cee/data-pred-ai/, sk https://industrialprocessai.com/sk/cee/data-pred-ai/, hu https://industrialprocessai.com/hu/cee/adatok-az-mi-elott/, ro https://industrialprocessai.com/ro/cee/datele-inaintea-ia/ HomeGuide · Data and AI · status 22 September 2026 ## Why only 3-8% of CEE manufacturers use AI, and the data layer that has to come first Eurostat's 2025 survey puts AI use among manufacturers in Romania, Poland and Hungary at less than half the EU rate. The fix is rarely a better model. It is reliable machine data, collected every shift. In shortIn 2025, 3.2% of manufacturers with 10+ employees in Romania, 7.7% in Poland and 7.8% in Hungary used at least one AI technology, against 17.3% in the EU (Eurostat). Industrial AI needs complete, timestamped machine states, counts and stop reasons. Build that first: capture automatically, add reasons, use it daily, then analyse, then apply AI to what remains. The TEEPTRAK approach Industrial AI pays where the machine data already exists. The examples on this page share one precondition: a reliable record of what each machine did, minute by minute. TEEPTRAK builds that record without touching the PLC and turns it into loss visibility the plant can act on the same week. - Run, stop and cycle data captured automatically on every machine - Stop reasons that give models labelled history, not only raw signals - Live OEE for operators, so the people who know the machines check data quality every shift Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Score your data readiness → ### The adoption numbers Enterprises using at least one AI technology, 10+ employees, % of enterprises | Country | All sectors 2024 | All sectors 2025 | Manufacturing 2024 | Manufacturing 2025 | | Romania | 3.1 | 5.2 | 1.3 | 3.2 | | Poland | 5.9 | 8.4 | 5.1 | 7.7 | | Hungary | 7.4 | 10.4 | 4.8 | 7.8 | | Czechia | 11.3 | 17.6 | 9.6 | 16.7 | | Slovakia | 10.8 | 18.0 | 8.1 | 15.8 | | EU-27 | 13.5 | 20.0 | 10.6 | 17.3 | | Germany (reference) | 19.8 | 26.0 | 16.1 | 24.4 | Source: Eurostat isoc_eb_ai (all sectors) and isoc_eb_ain2 (manufacturing, NACE C), indicator E_AI_TANY, 2024-2025. Eurostat broadened the AI questions in 2025, so part of the 2024→2025 increase is a survey change, not only real adoption. Two groups stand out. In Romania, Poland and Hungary, manufacturing AI use is under half the EU rate. In Czechia and Slovakia it roughly doubled in a year and is now close to the EU level, although the survey change explains part of that jump. Across the EU, the most common reason given by enterprises that considered AI but did not use it was lack of relevant expertise, cited by 70.9% of them, followed by unclear legal consequences (52.5%) and data-protection concerns (48.8%) (Eurostat, Use of artificial intelligence in enterprises, 2025). Eurostat does not publish a figure for data quality as a barrier in that release, so we do not quote one. ### Why industrial AI projects stall on data We have no reliable statistic on how many industrial AI projects fail, and we distrust the ones that circulate. The mechanism is easier to show than to count. A model that predicts a failure, recommends a setting or explains a loss needs history. That history must meet four conditions. - Complete. Every stop, including the short ones, with start and end times. A model trained on hand-logged breakdowns never sees the two-minute jams that make up much of the loss. - Labelled. Stops need reasons from a short, stable list. 'Other' as the largest category teaches a model nothing. - Aligned. Machine states, part counts, part numbers, shifts and, where available, process values must share one clock. Aligning a PLC log with a paper shift report after the fact is slow and often impossible. - Long and stable enough. Months of data, collected the same way, on a process that has not been redefined every quarter. If these conditions are not met, the first phase of an AI project becomes a data project, whether it was planned that way or not. It is cheaper to budget the data layer as its own project, with its own payback in daily loss reduction, and treat AI as a later option on top. The expertise gapThe 70.9% expertise barrier is also a data-layer argument. Stages 1 to 3 below need production and maintenance people, not data scientists. Stage 5 is the only one where specialist AI skills are essential, and by then the problem is narrow enough to buy in. ### A five-stage roadmap - Capture machine states and counts automaticallyEvery key machine reports running, stopped and producing, with part counts, from a signal rather than a pen. Done looks like: for any shift in the last month you can show every stop with its start and end, and the count matches production reporting within a small, known difference. - Put reasons on stopsOperators pick a reason from a short list, at the machine or on a tablet, while the stop is happening. Maintenance owns the list. Done looks like: more than nine in ten stop minutes have a reason, and 'other' is not in the top three. - Use it daily in shift meetingsThe start-of-shift meeting uses yesterday's losses by machine and reason, not memory. Actions are logged with owners. Done looks like: supervisors ask for the screen when it is missing, and the top loss on each key line has an open action. - Analytics on stable dataWith several months of consistent data, look at patterns: losses by part number, shift, changeover type, time since maintenance. This is where MTBF, MTTR and OEE trends become reliable enough for audits and customers. Done looks like: you can answer 'why did line 3 lose four points last month' in an hour, with numbers. - ML and AI on the problems that remainPick a narrow, costly problem that analytics did not solve: a failure mode with warning signs, a quality drift linked to process values, a scheduling problem. Add the sensors or data that problem needs. Done looks like: a model with a measured effect on one named loss, owned by someone in production. Stages 1 to 3 also produce the evidence that IATF 16949 clause 8.5.1.5 asks for on maintenance objectives; see what auditors expect on OEE, MTBF and MTTR. With energy sub-meters added, the same layer answers customer energy-per-part questions. Free data CEE Manufacturing Data Pack 2026 Every indicator on this page for 7 countries, EU-27 and Germany, with Eurostat dataset codes and links. Excel.Get the data pack → ### Checklist: is your data ready for AI? Answer for one key line, not for the plant | Check | Ready | Not ready yet | | How are stops recorded? | Automatically from the machine signal, with timestamps | On paper or typed in at the end of the shift | | Are short stops visible? | Yes, down to the minute or below | Only stops longer than a few minutes | | Share of stop time with a reason | Above about 90%, 'other' small | Large 'other' or 'unknown' category | | Part counts | From the machine, per part number | From the shift report or ERP postings | | Common clock | States, counts and process data share one time base | Separate systems, aligned by hand | | History | Six months or more collected the same way | Weeks, or definitions changed recently | | Daily use | Used in shift meetings; errors get noticed and fixed | Looked at monthly or for audits | | Ownership | Named owner for reason codes and definitions | Nobody, or IT only | Thresholds are practical rules of thumb, not published benchmarks. If most answers are in the right-hand column, an AI project will spend its first months fixing them. That work is worth doing, but call it what it is when you budget and when you apply for funding. ### How this links to funding Several CEE digitalisation programmes name production data collection, MES or IoT as eligible costs, which fits stages 1 to 3 well. Call status changes often. Our sister site lists the schemes by country with dates and official links: CEE factory digitalisation funding. Check the current call documentation with the agency; this is not grant advice. ### Where the data layer can come from Three common routes | Route | Fits when | Watch for | | MES | You also need orders, traceability and work instructions on the floor | Longer projects; machine-state capture on older machines may still need extra hardware | | Local integrator (PLC and SCADA data) | Machines have accessible controllers and you have in-house IT support | Definitions and dashboards are custom; keep the documentation | | Monitoring product (such as TEEPTRAK) | You want states, stop reasons and counts quickly, including on older machines | Subscription cost; integration with ERP or MES if you need it later | We are not neutral about the third route. All three can deliver stages 1 to 3. So can a disciplined spreadsheet on one line, as a first test of whether your teams will use the data. Free data CEE Manufacturing Data Pack 2026 Every indicator on this page for 7 countries, EU-27 and Germany, with Eurostat dataset codes and links. Excel.Get the data pack → - OEE system specification template for grant-funded projects - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5: OEE, MTBF and MTTR audit evidence - Answering OEM energy-per-part and OEE requests - IATF TPM Evidence Pack (xlsx) ### Questions **Why is AI use in Czech and Slovak manufacturing so much higher than in Poland or Romania?** The Eurostat data does not explain causes. Both countries roughly doubled between 2024 and 2025, and part of that jump comes from the broader 2025 questionnaire. Treat the gap as a signal of different starting points, not as a ranking of plants. **Can we skip stages and start with predictive maintenance?** You can, on one machine with a clear failure mode and dedicated sensors. Expect the project to rebuild parts of stages 1 and 2 for that machine. Plant-wide AI without stages 1 to 3 usually stalls on missing or unlabelled data. **How long does it take to get from stage 1 to stage 4?** It depends on how quickly reason codes and daily use settle. Analytics needs several months of data collected the same way. We do not publish a typical duration because we have no verified figure. **Do we need data scientists for stages 1 to 3?** No. These stages need production, maintenance and quality people who own the definitions and use the data every day. Specialist skills matter mainly at stage 5. **Does the data layer help with IATF audits?** Yes. Automatically captured stops and counts give a stronger basis for the OEE, MTBF and MTTR objectives that IATF 16949 clause 8.5.1.5 names as examples, and for their monthly trend and review. ### Sources - Eurostat isoc_eb_ai, enterprises using AI technologies (2024-2025) - Eurostat isoc_eb_ain2, AI use by NACE Rev.2 activity, manufacturing (2024-2025) - Eurostat Statistics Explained, Use of artificial intelligence in enterprises (2025 data) Figures are sourced on each page. Funding rules change: check the official call documents before you budget. --- # IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence URL: https://industrialprocessai.com/cee/iatf-16949-tpm-oee/ Summary: What IATF 16949 clause 8.5.1.5 asks for on total productive maintenance, what evidence auditors look at for OEE, MTBF and MTTR, and where Excel records fail. Translations: pl https://industrialprocessai.com/pl/cee/iatf-16949-tpm-oee/, cs https://industrialprocessai.com/cs/cee/iatf-16949-tpm-oee/, sk https://industrialprocessai.com/sk/cee/iatf-16949-tpm-oee/, hu https://industrialprocessai.com/hu/cee/iatf-16949-tpm-oee/, ro https://industrialprocessai.com/ro/cee/iatf-16949-tpm-oee/ HomeCompliance · IATF 16949 · status 22 September 2026 ## IATF 16949 §8.5.1.5 total productive maintenance: what auditors expect on OEE, MTBF and MTTR For quality, maintenance and plant managers at tier-1 and tier-2 automotive suppliers. What the clause asks for in plain terms, what evidence tends to be checked, and why the source of your downtime data matters. In shortIATF 16949 clause 8.5.1.5 requires a documented total productive maintenance system: critical equipment identified, spare parts and maintenance resources in place, preventive and predictive methods, periodic overhaul, and documented maintenance objectives such as OEE, MTBF and MTTR that are reviewed and acted on when missed. Auditors check the objectives per critical machine, the data behind them, and the actions taken. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → ### What is verified on this page and what is practice The IATF 16949:2016 standard is sold, not published. We have not quoted it. The clause content below is paraphrased from a published clause summary and from the IATF's own public FAQ document, both listed in the sources. The IATF FAQ confirms that clause 8.5.1.5 contains a list of specific requirements, items a) to j), and that these overlap with, but are not the same as, the TPM pillars used in industry. The sections on audit evidence and weak points are general audit practice, not IATF text. Buy the standard from IATF, VDA-QMC or AIAG before you build your system on it. Why this clauseIn the ranking of IATF 16949 minor nonconformities published with data as of 25 November 2024, clause 8.5.1.5 was the most frequent minor finding, at 4.30% of minor nonconformities (simpleQuE summary of IATF data; the IATF Global Oversight top-ten document is listed in sources). It did not appear in the top ten major findings in that summary. ### What the clause requires, in plain terms Paraphrased from the clause summary we used. Check each line against your licensed copy of the standard. - A documented TPM system. Not a slogan and not only a preventive-maintenance calendar. A described system with responsibilities, methods and records. - Identification of key process equipment. The machines, tools and gauges you need to make conforming product at the required volume. This list is the scope of everything else. - Availability of replacement parts for that key equipment, so a failure does not become a long stop. - Resources for maintenance: people, skills, tools and budget. - Packaging and preservation of equipment, tooling and gauging, including items in storage or idle between runs. - Customer-specific requirements that apply to maintenance. - Documented maintenance objectives, with OEE, MTBF and MTTR named as examples. - Regular review of the maintenance plan and objectives, with a documented action plan when objectives are not met. Clause summaries link this review to management review. - Preventive maintenance methods, and predictive maintenance methods where applicable. - Periodic overhaul. The IATF FAQ explains this as major maintenance, such as disassembly, repair, part replacement and reassembly, done proactively when routine upkeep is not enough. Some tooling is replaced rather than overhauled; the equipment maker's guidance helps set intervals. Two practical consequences. First, the objectives are expected per key equipment, not only as a plant average. Second, an objective without a review record and an action plan when it is missed is half a requirement. ### What an auditor typically asks to see This is common audit practice, not a list from the standard. Expect the auditor to pick one or two critical machines from your control plan or your key-equipment list and follow the thread. - The key-equipment list and how a machine gets on it. Machines that feed a customer-critical characteristic and have no objectives are an easy finding. - The objectives and targets for that machine: OEE, MTBF, MTTR, and often preventive-maintenance completion. Targets should have a reason, such as capacity needed for the customer's volumes. - A trend over time, usually monthly for at least the last year. One figure on a slide does not show control. - How the numbers are produced: who records stops, when, with which reason codes, and how the calculation is done. Auditors ask the operator and the maintenance technician, not only the quality manager. - The review record: minutes or a management-review input showing the objectives were looked at. - Actions when a target is missed, with owners and dates, and evidence that the action changed the trend. - Links to the rest of the system: spare-parts stock for the top failure modes, preventive-maintenance plans updated after repeated failures, and the contingency plan for that machine. Free tool IATF TPM Evidence Pack OEE, MTBF and MTTR objectives tracker per machine, management-review summary, audit evidence checklist and an energy-per-good-part sheet. Excel.Get the evidence pack → ### Objective, calculation, data needed and common finding Short versions of the usual formulas. For full definitions and variants, see manufacturing-metrics.org. The findings column is typical audit experience, not IATF statistics. TPM objectives under IATF 16949 §8.5.1.5: what each one needs | Objective | How it is usually calculated | Data you need | Common audit finding | | OEE | Availability × performance × quality, over planned production time | Planned time, run time, every stop with start and end, ideal cycle time per part number, good and total counts | Plant-level OEE only; no figure per key machine. Ideal cycle time changed quietly to make performance look better. | | MTBF | Operating time ÷ number of failures in the period | Operating time per machine, a clear rule for what counts as a failure, failure events with timestamps | No written definition of 'failure'. Short stops left out, so MTBF looks far better than the line feels. | | MTTR | Total repair time ÷ number of repairs | Start and end of each breakdown, ideally split into waiting for technician, diagnosis, waiting for parts and repair | Repair time starts when maintenance arrives, not when the machine stopped. Waiting for parts is invisible. | | Preventive-maintenance completion (common addition) | PM tasks done on time ÷ PM tasks due | PM plan per machine, completion dates, reasons for postponement | Tasks closed in the system with no record of what was done. Overdue tasks rolled forward without review. | | Spare-parts availability (supporting) | Breakdowns delayed by missing parts ÷ all breakdowns | Breakdown records with a 'waiting for parts' reason, stock list for critical spares | Critical-spares list exists but was never compared with the actual failure history. | Formulas are standard industry practice. The clause names OEE, MTBF and MTTR as examples of objectives; it does not prescribe these formulas. ### Where manual and Excel-based objectives break down Many suppliers run §8.5.1.5 objectives from a paper stop sheet typed into a spreadsheet. It can pass an audit. It is also where most weak evidence comes from, for four reasons. - Short stops are not written down. An operator will log a 40-minute breakdown. A two-minute jam repeated thirty times a shift rarely makes it to paper. Availability looks fine and performance loss is unexplained. - Times are rounded and entered after the fact. 'Stopped 10:00-10:30' written at the end of the shift is an estimate. MTTR built on estimates cannot show a real improvement of a few minutes. - Reasons drift. 'Other' and 'technical' grow until they are the largest categories. An auditor who sees 'other' as the top loss will ask what you did about it. - The file can be edited without a trace. A spreadsheet with no change log makes it hard to show that last March's figures are what they were last March. Automatically captured machine states fix the first two points: every stop has a start and end time from the machine signal, including the short ones, and the operator only adds the reason. The record is created when the event happens, not when someone has time. That makes the trend more believable to an auditor and more useful to your maintenance team. It does not fix the third point on its own; reason codes still need an owner and a short list. Automatic capture can come from your MES, from PLC data collected by a local integrator, or from a production-monitoring product. Any of the three works for this clause if the data is complete and the definitions are written down. ### For suppliers in Czechia, Slovakia, Hungary, Romania and Poland Automotive is the largest manufacturing sub-sector by gross value added in Czechia, Hungary and Slovakia, and the second largest in Romania (Eurostat nama_10_a64, 2023). That puts a large share of the region's factories under IATF 16949 audits, directly or through their customers. Automotive weight in the five countries | Country | Figure | Source and year | | Czechia | 1.47 million vehicles produced in 2025; 136,956 employees at AutoSAP member companies | AutoSAP via Kurzy.cz, 2026 | | Slovakia | About 1.07 million cars in 2025, 196 per 1,000 inhabitants; ZAP forecasts 1.02 million in 2026 | ZAP via AutoFoxNews, 2026; The Slovak Spectator, Jan 2026 | | Hungary | About 150,000 employed; almost a fifth of industrial output | AutoSAP, 2025 | | Romania | More than 220,000 employees in more than 630 companies; 13.2% of GDP and 33% of exports (industry-association figures, not cross-checked with INS) | The Bridge of Friendship citing industry data, Apr 2026 | | Poland | 197,000 employed in automotive (Sept 2024); about 8% of GDP | PAIH, 2025 | Figures come from different bodies with different definitions. Do not add them up. Two local points matter for TPM. Manufacturing labour costs rose 29.5% in Poland and 25.0% in Romania between 2023 and 2025 (Eurostat lc_lci_lev), and industry bodies in Czechia and Slovakia name skilled-worker shortages as a limit on growth (AutoSAP 2026, ZAP 2026). Maintenance technicians are part of that shortage. Objectives that show where their hours go, such as MTTR split into waiting and repair time, help you argue for the people and spares you need, which is also what §8.5.1.5 asks you to provide. ### A short plan before the next surveillance audit - Fix the key-equipment listStart from the control plan. Every machine that carries a special characteristic or is a capacity bottleneck goes on it. - Write the definitionsOne page: what a failure is, when repair time starts and stops, which stops count as planned, and where ideal cycle times come from. - Set targets per machineBase them on the last twelve months and on the capacity your customers need. Record why each target was chosen. - Put the review in the calendarMonthly at maintenance level, with a summary into management review. Keep minutes and an action list with owners and dates. - Close the data gapsWherever stops are hand-logged on a key machine, decide whether to capture machine state automatically. Start with the machines behind your worst objectives. Free tool IATF TPM Evidence Pack OEE, MTBF and MTTR objectives tracker per machine, management-review summary, audit evidence checklist and an energy-per-good-part sheet. Excel.Get the evidence pack → - OEE system specification template for grant-funded projects - What an OEE system really costs: the 3-year calculator - IATF TPM Evidence Pack (xlsx) - Answering OEM energy-per-part and OEE data requests - Why the data layer has to come before AI ### Questions **Does IATF 16949 require us to measure OEE?** The clause asks for documented maintenance objectives and names OEE, MTBF and MTTR as examples, according to published summaries of clause 8.5.1.5. Most auditors expect to see these or a justified equivalent for key equipment. Check the wording in your licensed copy of the standard and any customer-specific requirements. **Do we need maintenance objectives for every machine?** The clause is built around key process equipment, the equipment needed to make conforming product at the required volume. Define that list with a clear rule, and set objectives for the machines on it. Machines outside the list still need maintenance, but objectives per machine are expected mainly for the key ones. **Is an Excel file acceptable evidence?** Yes, if the data is complete, the definitions are written, and the review and actions are recorded. The weak points are missing short stops, times estimated after the fact and edits without a trace. Auditors test these by asking operators how stops are recorded. **What does periodic overhaul mean?** The IATF FAQ describes it as major maintenance done proactively when routine upkeep is not enough, such as disassembly, repair, part replacement and reassembly. Some tooling is simply replaced instead. Use the equipment maker's recommendations to set intervals. **How often should maintenance objectives be reviewed?** The clause requires regular review with an action plan when objectives are missed; it does not set a frequency in the summaries we used. Monthly review at maintenance level with a summary into management review is common practice and easy to show at audit. ### Sources - IATF 16949:2016 Frequently Asked Questions (FAQs 26 and 27 on clause 8.5.1.5), IATF, Oct 2019 - Clause 8.5.1.5 summary, Pretesh Biswas (secondary source), 2023 - IATF Global Oversight, top-ten minor nonconformities, 8.5.1.5 (Dec 2024) - simpleQuE, Top 10 IATF 16949 major and minor nonconformances of 2024 - Eurostat nama_10_a64, gross value added by industry (2023) - Eurostat lc_lci_lev, labour cost levels (2023-2025) - AutoSAP via Kurzy.cz, Czech vehicle production 2025 (2026) - ZAP via AutoFoxNews, Slovak car production 2025 (2026) - The Slovak Spectator, ZAP outlook (13 Jan 2026) - AutoSAP, Hungary automotive (2025) - The Bridge of Friendship, Romanian automotive industry (Apr 2026) - PAIH, The Automotive & Electromobility Sector 2025 Figures are sourced on each page. Funding rules change: check the official call documents before you budget. --- # IATF TPM Evidence Pack: OEE, MTBF, MTTR tracker (xlsx) URL: https://industrialprocessai.com/cee/iatf-tpm-evidence-pack/ Summary: An xlsx for IATF 16949 §8.5.1.5: per-machine OEE, MTBF and MTTR objectives, management-review summary, audit evidence checklist and energy-per-part sheet. Translations: pl https://industrialprocessai.com/pl/cee/pakiet-dowodow-iatf-tpm/, cs https://industrialprocessai.com/cs/cee/balicek-dukazu-tpm-iatf/, sk https://industrialprocessai.com/sk/cee/balik-dokazov-iatf-tpm/, hu https://industrialprocessai.com/hu/cee/iatf-tpm-bizonyitek-csomag/, ro https://industrialprocessai.com/ro/cee/pachet-dovezi-tpm-iatf/ HomeDownload · IATF 16949 · status 22 September 2026 ## IATF TPM Evidence Pack A spreadsheet for keeping total productive maintenance objectives, reviews and audit evidence in one place. In shortThe IATF TPM Evidence Pack is an xlsx workbook for automotive suppliers working to IATF 16949 clause 8.5.1.5. It holds per-machine OEE, MTBF and MTTR targets with monthly actuals, a management-review summary, an audit evidence checklist written in paraphrase, and an energy-per-good-part calculator. It organises your evidence; it does not certify anything. What is inside the workbook: ### Your file is ready The download starts below. We have also emailed you the link. Download the file - Objectives tracker per machine. One row per key equipment item: OEE, MTBF and MTTR targets, the reason for each target, and monthly actuals with a twelve-month trend. - Management-review summary sheet. Pulls the machines that missed target, the actions, owners and due dates onto one page you can attach to management-review minutes. - Audit evidence checklist. The items of clause 8.5.1.5 in our own paraphrase, each with the record that usually serves as evidence and a column for where yours is kept. - Energy-per-good-part calculator. Energy by machine state, good parts and the idle-energy cost, for customer questionnaires. - Notes sheet. Definitions to fill in: what counts as a failure, when repair time starts and stops, how ideal cycle time is set. Who it is for. Quality, maintenance and plant managers at tier-1 and tier-2 suppliers, in Czechia, Slovakia, Hungary, Romania, Poland or elsewhere, who keep TPM objectives in Excel today or want to check what their current system is missing before a surveillance audit. LimitsIt is not a certification tool and does not replace the standard. IATF 16949:2016 is paywalled; buy it from IATF, VDA-QMC or AIAG and check every requirement against your licensed copy and your customer-specific requirements. The workbook also cannot fix hand-logged data: if short stops are not recorded, the MTBF it shows will be too optimistic. TEEPTRAK's CEE team in Bucharest may follow up once after you download it. ### Sources - IATF 16949:2016 Frequently Asked Questions, IATF, Oct 2019 Figures are sourced on each page. Funding rules change: check the official call documents before you budget. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → --- # OEM supplier data requests: energy per part and OEE after Omnibus URL: https://industrialprocessai.com/cee/oem-supplier-data-requests/ Summary: What changed for CEE suppliers with the 2026 Omnibus I, what customers will still ask for, and how to calculate energy per good part from machine-state data. Translations: pl https://industrialprocessai.com/pl/cee/zapytania-oem-dane-dostawcow/, cs https://industrialprocessai.com/cs/cee/pozadavky-oem-na-data-dodavatelu/, sk https://industrialprocessai.com/sk/cee/datove-poziadavky-oem/, hu https://industrialprocessai.com/hu/cee/oem-beszallitoi-adatkeresek/, ro https://industrialprocessai.com/ro/cee/solicitari-date-oem-furnizori/ HomeCompliance · Customer data requests · status 22 September 2026 ## Your customer wants energy-per-part and OEE data: answering OEM supplier questionnaires after the 2026 Omnibus Omnibus I cut the number of companies that report under the CSRD and capped what they can ask smaller suppliers. The questionnaires have not stopped. This page covers what to have ready and how to calculate it. In shortSince Omnibus I entered into force on 18 March 2026, the CSRD applies to companies with more than 1,000 employees and €450m turnover, and they may not ask suppliers under 1,000 employees for more sustainability information than a voluntary standard covers. Expect requests for company energy and emissions, plus commercial questions on delivery, quality and capacity. Energy per good part needs machine-state and count data. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → ### What changed with Omnibus I Status as of 22 September 2026. This is a summary for operations and quality managers, not legal advice. Check the national transposition in your country. Omnibus I: the points that matter for a CEE supplier | Item | What it says | Source | | Publication and entry into force | Published in the Official Journal on 26 Feb 2026; in force since 18 Mar 2026 | Latham & Watkins, 27 Feb 2026 | | CSRD scope | EU companies or groups with more than 1,000 employees and more than €450m net turnover | Latham & Watkins, 2026 | | Value-chain cap | Limits the sustainability information reporting companies may request from value-chain companies with fewer than 1,000 employees | Latham & Watkins, 2026 | | What sets the limit | A voluntary reporting standard for companies protected by the cap, derived from the Commission's VSME Recommendation (EU) 2025/1710. The Commission adopted it as a delegated act on 3 Jul 2026, subject to Council and Parliament scrutiny | Linklaters, 2026 | | When the cap applies | From financial year 2027, per Linklaters' summary | Linklaters, 2026 | | Transposition | Member States must transpose the CSRD changes by 19 Mar 2027 | Latham & Watkins, 2026 | | CSDDD (due diligence) | Scope more than 5,000 employees and €1.5bn turnover; transposition by 26 Jul 2028 | Latham & Watkins, 2026 | Check whether the voluntary-standard delegated act has been published in the Official Journal before relying on its exact content. For most plants in Czechia, Slovakia, Hungary, Romania and Poland, this means two things. Few will report under the CSRD themselves. Many will still receive requests from Western customers who do, but those requests should be narrower and more standardised than the long questionnaires of 2024. ### What a CEE supplier will still be asked The voluntary SME standard (VSME) basic module includes disclosure B3 on energy and greenhouse gas emissions: total energy consumption in MWh, split by renewable and non-renewable and by electricity and fuels; gross Scope 1 and location-based Scope 2 emissions; and GHG intensity per unit of turnover (EFRAG Knowledge Hub, VSME Annex I). If you are under 1,000 employees, expect a customer's CSRD-driven request to look like this. It is company-level data. It is not per part. Customers also ask for things that have nothing to do with the CSRD: delivery performance, quality (ppm, complaints), capacity and OEE on lines that run their parts, and sometimes energy or carbon per part for product-level footprints. The cap is written around sustainability information requested for reporting. Whether a purchasing or product-footprint question falls under it is a legal question. Ask the customer which basis the request has, and get advice if the answer matters to you. Practical readingThe company-level energy figure comes from your bills. The per-part figure, and anything that explains it, comes from the shop floor. Only the second one tells you or your customer where the energy goes. ### Why energy per good part needs machine-state data Dividing the plant's monthly kWh by parts shipped gives a number. It mixes lighting, compressed air, heating and every product family. It moves when the product mix moves. It cannot tell you whether an improvement came from better machines or from a lucky month. A useful energy-per-part figure needs three things per machine or line: energy consumed (a meter or a sub-meter), the time spent in each state (producing, idle, stopped, changeover), and good-part counts. With those you can separate energy used making good parts from energy used while the machine is idle, making scrap or waiting. Idle and scrap energy is where suppliers usually find something to fix. It is also the same state and count data that feeds OEE, so one data layer serves both the maintenance objectives in IATF 16949 §8.5.1.5 and the customer's energy question. ### Industrial electricity prices in the region Electricity, non-household consumers, 2,000-19,999 MWh/year, excluding VAT and recoverable taxes, 2025 second half | Country | EUR per kWh | vs EU-27 average | | Hungary | 0.234 | +0.042 | | Slovakia | 0.219 | +0.027 | | Poland | 0.214 | +0.022 | | Romania | 0.207 | +0.015 | | Czechia | 0.198 | +0.006 | | EU-27 | 0.192 | · | | Germany (reference) | 0.237 | +0.045 | Source: Eurostat nrg_pc_205, band MWH2000-19999, I_TAX, EUR, 2025-S2. Differences are our calculation. Mid-size plants in all five countries paid more than the EU average in the second half of 2025, and Hungary and Slovakia were close to German levels. For a plant in this band, idle-machine energy is a cost line, not only a reporting item. ### The answer kit: numbers to have ready What to prepare, at what granularity | Number | Granularity | Where it comes from | Who usually asks | | Total energy consumption (MWh), electricity and fuels, renewable share | Company or site, yearly | Utility bills, supplier certificates | CSRD-reporting customers (VSME B3 scope) | | Scope 1 and location-based Scope 2 emissions (tCO2e) | Company or site, yearly | Fuel and electricity data with emission factors | CSRD-reporting customers | | GHG intensity per turnover | Company, yearly | Emissions ÷ turnover | CSRD-reporting customers | | Energy per good part | Per line or part family, monthly | Sub-meters, machine states, good counts | Customer purchasing or product-footprint teams | | OEE and its three components | Per line that runs the customer's parts, monthly | Machine states, cycle times, counts | Supplier quality, capacity reviews | | Delivery and quality performance | Per customer, monthly | ERP, quality system | Supplier quality | The first three rows follow the VSME basic module disclosure B3 (Recommendation (EU) 2025/1710). The other rows are common commercial requests, not regulatory requirements. ### How to calculate energy per good part Energy per good part (kWh/part) = energy consumed by the line in the period ÷ good parts produced in the period. Then split the energy by machine state. Productive energy share = kWh while producing ÷ total kWh. The rest is idle, stopped, changeover or start-up energy. Scrap energy ≈ kWh while producing × (scrap parts ÷ total parts). Illustrative numbers, not data from any plant: a press line uses 42,000 kWh in a month and makes 150,000 good parts, which is 0.28 kWh per good part. If 30% of the kWh was consumed while the line was idle or stopped, 12,600 kWh bought no parts. At the Polish 2025-S2 band price of €0.214/kWh, that is about €2,700 for the month on one line. - Use the same period and the same good-count definition as your OEE, so the two figures agree. - Allocate shared utilities (compressed air, chillers) by a written rule, or report them separately. Do not hide them in one line's figure. - Report per part family, not per plant, when the customer only buys one family. Free tool IATF TPM Evidence Pack OEE, MTBF and MTTR objectives tracker per machine, management-review summary, audit evidence checklist and an energy-per-good-part sheet. Excel.Get the evidence pack → ### What not to over-promise - Do not send per-part figures you cannot reproduce. If the number came from a plant total divided by shipments, say so. Customers compare year to year. - Do not promise real-time data feeds to a customer portal unless you have the capture, the definitions and someone to maintain them. - Do not answer more than was asked. If you are under 1,000 employees and the request is CSRD-driven, the voluntary standard sets the limit of what the customer may request. Answering beyond it is your choice. - Do not commit to reduction targets before you have twelve months of baseline at the granularity you will be measured on. - Do not treat OEE shared with a customer as an internal target. Agree the definition in writing first, especially planned downtime and ideal cycle time. The machine-state and count data described here can equally come from an MES, from PLC data collected by an integrator, or from sub-meters logged in a spreadsheet as a first step. Free tool IATF TPM Evidence Pack OEE, MTBF and MTTR objectives tracker per machine, management-review summary, audit evidence checklist and an energy-per-good-part sheet. Excel.Get the evidence pack → - OEE system specification template for grant-funded projects - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5: OEE, MTBF and MTTR audit evidence - IATF TPM Evidence Pack, with energy-per-part sheet - Why the data layer comes before AI ### Questions **Does the CSRD apply to my plant after Omnibus I?** Only if your company or group has more than 1,000 employees and more than €450m net turnover, under the thresholds published on 26 February 2026. National transposition is due by 19 March 2027. Most CEE supplier plants will fall below these thresholds, but their customers may not. **Can my customer still ask me for sustainability data if we have fewer than 1,000 employees?** Yes, within limits. The value-chain cap restricts CSRD-reporting companies to the information covered by the voluntary standard derived from the VSME Recommendation. Its basic module includes company energy use and Scope 1 and 2 emissions. Commercial requests may follow other rules; ask the customer which basis applies. **Is energy per part required by the VSME standard?** No. Disclosure B3 of the basic module covers total energy consumption, Scope 1 and location-based Scope 2 emissions and GHG intensity per turnover, all at company level. Per-part figures usually come from customer purchasing or product-footprint requests. **Can I calculate energy per part without sub-meters?** You can allocate the plant's energy by machine hours or rated power, but the result is an estimate and should be labelled as one. For a figure you can defend and improve, meter the main consumers and record machine states and good counts for the same period. **Should we share OEE with our customer?** Many suppliers do, especially for lines dedicated to a customer's parts. Agree the definitions first, including planned downtime and ideal cycle time, and share monthly figures you can reproduce rather than live screens. ### Sources - Latham & Watkins, EU Sustainability Omnibus published in the Official Journal (27 Feb 2026) - Linklaters, Commission adopts revised ESRS and voluntary reporting standard (2026) - EFRAG Knowledge Hub, VSME Standard (Annex I to Recommendation (EU) 2025/1710) - Eurostat nrg_pc_205, electricity prices for non-household consumers (2025-S2) Figures are sourced on each page. Funding rules change: check the official call documents before you budget. --- # Industrial AI Data Readiness Assessment URL: https://industrialprocessai.com/data-readiness/ Summary: Twenty-three checks across signal coverage, sampling, naming, history and governance: weighted, with a verdict at the bottom. Most failed pilots fail here, not in the model. Translations: it https://industrialprocessai.com/it/data-readiness/, pt https://industrialprocessai.com/pt/data-readiness/, pl https://industrialprocessai.com/pl/data-readiness/, cs https://industrialprocessai.com/cs/data-readiness/, sk https://industrialprocessai.com/sk/data-readiness/, hu https://industrialprocessai.com/hu/data-readiness/, sv https://industrialprocessai.com/sv/data-readiness/, tr https://industrialprocessai.com/tr/data-readiness/ Free tool ## Score your data before you score the vendors Twenty-three checks across signal coverage, sampling, naming, history and governance: weighted, with a verdict at the bottom. Most failed pilots fail here, not in the model. New: What an OEE system really costs: the 3-year calculator Done: your download is ready. Download the Assessment (.xlsx) Pair it with the published rubric. #### What’s inside - Twenty-three weighted checks, scored 0-3 - Five areas: coverage, sampling, naming, history, governance - Each check says why a model needs it - A readiness percentage and a plain verdict - Built to be filled in during a plant walk The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → --- # Industrial AI Guides URL: https://industrialprocessai.com/guides/ Summary: Plain-language guides to industrial AI and machine learning in manufacturing: definitions, examples, generative AI, and how to buy. Translations: fr https://industrialprocessai.com/fr/guides/, de https://industrialprocessai.com/de/guides/, es https://industrialprocessai.com/es/guides/, it https://industrialprocessai.com/it/guides/, nl https://industrialprocessai.com/nl/guides/, pt https://industrialprocessai.com/pt/guides/, pl https://industrialprocessai.com/pl/guides/, cs https://industrialprocessai.com/cs/guides/, sk https://industrialprocessai.com/sk/guides/, hu https://industrialprocessai.com/hu/guides/, ro https://industrialprocessai.com/ro/guides/, sv https://industrialprocessai.com/sv/guides/, tr https://industrialprocessai.com/tr/guides/, zh https://industrialprocessai.com/cn/guides/ Learn ## Guides Answer-first industrial-AI guides, written for engineers. The TEEPTRAK approach Industrial AI pays where the machine data already exists. The examples on this page share one precondition: a reliable record of what each machine did, minute by minute. TEEPTRAK builds that record without touching the PLC and turns it into loss visibility the plant can act on the same week. - Run, stop and cycle data captured automatically on every machine - Stop reasons that give models labelled history, not only raw signals - Live OEE for operators, so the people who know the machines check data quality every shift Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Score your data readiness → What is an industrial AI platform? A plain-language definition of industrial AI platforms, the four platform categories, what they cost to deploy in time and people,… AI in manufacturing: 12 examples that actually produce ROI Concrete 2026 examples of AI and machine learning in manufacturing: scrap reduction, predictive maintenance, energy optimization,… Generative AI in manufacturing: separating the layer that works from the hype Where generative AI genuinely helps manufacturers in 2026, and where industrial machine learning, not LLMs, does the real process… Before you shortlist Most failed pilots fail on data, not on the model Free, no login. One follow-up from TEEPTRAK, then nothing.Score your data readiness → --- # 12 Real AI in Manufacturing Examples That Produce ROI (2026) URL: https://industrialprocessai.com/guides/ai-manufacturing-examples/ Summary: Concrete 2026 examples of AI and machine learning in manufacturing (scrap reduction, predictive maintenance, energy optimization, visual inspection) with the outcome numbers that make the business case. Translations: fr https://industrialprocessai.com/fr/guides/ai-manufacturing-examples/, de https://industrialprocessai.com/de/guides/ai-manufacturing-examples/, es https://industrialprocessai.com/es/guides/ai-manufacturing-examples/, it https://industrialprocessai.com/it/guides/ai-manufacturing-examples/, nl https://industrialprocessai.com/nl/guides/ai-manufacturing-examples/, pt https://industrialprocessai.com/pt/guides/ai-manufacturing-examples/, pl https://industrialprocessai.com/pl/guides/ai-manufacturing-examples/, cs https://industrialprocessai.com/cs/guides/ai-manufacturing-examples/, sk https://industrialprocessai.com/sk/guides/ai-manufacturing-examples/, hu https://industrialprocessai.com/hu/guides/ai-manufacturing-examples/, ro https://industrialprocessai.com/ro/guides/ai-manufacturing-examples/, sv https://industrialprocessai.com/sv/guides/ai-manufacturing-examples/, tr https://industrialprocessai.com/tr/guides/ai-manufacturing-examples/, zh https://industrialprocessai.com/cn/guides/ai-manufacturing-examples/ Home / Guides Guide · updated September 1, 2026 ## AI in manufacturing: 12 examples that actually produce ROI The short answerThe AI-in-manufacturing examples producing measurable ROI in 2026 cluster into four families: quality & scrap (root-cause ML, up to −58% scrap in published deployments), equipment (predictive maintenance cutting unplanned downtime 20-50%), resources (energy optimization, −10 to −20%), and inspection (vision systems catching defects humans miss at line speed). The common thread: models trained on the plant's own data, operated by its own engineers. The TEEPTRAK approach Industrial AI pays where the machine data already exists. The examples on this page share one precondition: a reliable record of what each machine did, minute by minute. TEEPTRAK builds that record without touching the PLC and turns it into loss visibility the plant can act on the same week. - Run, stop and cycle data captured automatically on every machine - Stop reasons that give models labelled history, not only raw signals - Live OEE for operators, so the people who know the machines check data quality every shift Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Score your data readiness → ### Quality & scrap 1 · Unsupervised root-cause discovery. ML learns normal behavior across hundreds of process variables and traces scrap spikes to their true causes: material-lot interactions, drift combinations, sequence effects. Published deployments report scrap reductions of 10-30%, with higher figures reported by vendors about their own sites. 2 · Predictive quality: models score in-process product against final quality, catching bad runs mid-run. 3 · Golden-run replication: platforms mine the best historical runs and recommend the settings that reproduce them shift after shift. ### Equipment 4 · Vibration-based failure prediction on rotating equipment, with prescriptive diagnostics. 5 · Process-signal drift detection (failure precursors that appear in process variables before any vibration change. 6 · Maintenance scheduling optimization) moving from calendar-based to condition-based intervals, typically recovering 20-50% of unplanned downtime. ### Resources 7 · Energy optimization: ML finds the settings and sequences that cut kWh per unit: published results around −20% energy on optimized lines. 8 · Utility and compressed-air anomaly detection, the classic invisible leak. 9 · Raw-material yield optimization in recipe-driven industries. ### Inspection & frontline 10 · Automated visual inspection: vision models at line speed with defect classification. 11 · Digitized quality checks with anomaly flags at the workstation. 12 · Generative-AI copilots that explain alerts, draft root-cause reports and query the historian in plain language: the newest layer, augmenting rather than replacing the industrial ML underneath. ### What the winners have in common None of these examples came from a moonshot data-science project. Every one runs on the plant's existing data, was live within days or weeks, and is operated by the plant's own engineers. That deployment profile (not the algorithm) is what separates the ROI stories from the pilot graveyard. ### FAQ **What are the best examples of AI in manufacturing?** The highest-ROI examples in 2026: ML root-cause analysis of scrap and quality drift, predictive maintenance on critical equipment, energy optimization across process lines, automated visual inspection, and golden-run setpoint recommendation for continuous processes. **How can generative AI help in manufacturing?** Generative AI helps with the language side of the factory: writing and querying work instructions, summarizing shift handovers, explaining alerts in plain language, and accelerating root-cause reports. The process math itself still comes from industrial ML models trained on plant data; the two are complementary layers. **Will AI replace manufacturing jobs?** The evidence so far points to task change rather than headcount replacement: AI removes threshold-babysitting and manual data wrangling while creating demand for engineers who can act on model output. Plants deploying process AI typically redeploy attention to improvement work rather than cutting operators. Before you shortlist Most failed pilots fail on data, not on the model Free, no login. One follow-up from TEEPTRAK, then nothing.Score your data readiness → --- # Generative AI in Manufacturing: What Works, What Doesn't (2026) URL: https://industrialprocessai.com/guides/generative-ai-manufacturing/ Summary: Where generative AI genuinely helps manufacturers in 2026, and where industrial machine learning, not LLMs, does the real process work. Use cases, architecture, cautions. Translations: fr https://industrialprocessai.com/fr/guides/generative-ai-manufacturing/, de https://industrialprocessai.com/de/guides/generative-ai-manufacturing/, es https://industrialprocessai.com/es/guides/generative-ai-manufacturing/, it https://industrialprocessai.com/it/guides/generative-ai-manufacturing/, nl https://industrialprocessai.com/nl/guides/generative-ai-manufacturing/, pt https://industrialprocessai.com/pt/guides/generative-ai-manufacturing/, pl https://industrialprocessai.com/pl/guides/generative-ai-manufacturing/, cs https://industrialprocessai.com/cs/guides/generative-ai-manufacturing/, sk https://industrialprocessai.com/sk/guides/generative-ai-manufacturing/, hu https://industrialprocessai.com/hu/guides/generative-ai-manufacturing/, ro https://industrialprocessai.com/ro/guides/generative-ai-manufacturing/, sv https://industrialprocessai.com/sv/guides/generative-ai-manufacturing/, tr https://industrialprocessai.com/tr/guides/generative-ai-manufacturing/, zh https://industrialprocessai.com/cn/guides/generative-ai-manufacturing/ Home / Guides Guide · updated September 1, 2026 ## Generative AI in manufacturing: separating the layer that works from the hype The short answerGenerative AI's real manufacturing role in 2026 is the language layer: explaining alerts, drafting root-cause and shift reports, querying plant data conversationally, and maintaining work instructions. The process intelligence itself (anomaly detection, root-cause discovery, setpoint optimization) still comes from industrial ML trained on the plant's own time-series data. The winning architecture pairs both; the failing projects ask an LLM to do a process model's job. The TEEPTRAK approach Industrial AI pays where the machine data already exists. The examples on this page share one precondition: a reliable record of what each machine did, minute by minute. TEEPTRAK builds that record without touching the PLC and turns it into loss visibility the plant can act on the same week. - Run, stop and cycle data captured automatically on every machine - Stop reasons that give models labelled history, not only raw signals - Live OEE for operators, so the people who know the machines check data quality every shift Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Score your data readiness → ### The architecture that works Every credible 2026 deployment we reviewed uses the same two-layer pattern. Layer 1 (industrial ML: models trained on the plant's own process history (unsupervised process-ML platforms, machine-health models, OEE and machine-data systems) do the numerical work) anomaly detection across hundreds of variables, root-cause attribution, failure prediction. Layer 2 (generative AI: LLMs translate that output into human decisions) "explain this alert to the night shift", "draft the 8D report", "what changed on line 3 since Tuesday?". ### Five uses that pay today 1 · Alert explanation. The gap between "model flagged an anomaly" and "operator acts" is language; LLMs close it. 2 · Root-cause reports drafted in minutes from model output plus context. 3 · Conversational plant data (historian queries without SQL. 4 · Work instructions generated and kept current from process changes. 5 · Shift handover and maintenance-log summarization) unglamorous, universally adopted once tried. ### Where it fails Projects that ask an LLM to be the process model (predict quality from raw sensor streams, set process parameters, replace an historian) fail on accuracy, latency and trust. Hallucinated confidence is harmless in a marketing draft and dangerous next to a reactor. Guardrail rule: generative AI may explain and draft; only validated industrial ML output and humans change the process. ### What to buy, in what order Data capture first (see our), industrial ML second, generative layer third: increasingly arriving built into the platforms themselves rather than as a separate purchase. Evaluate any "AI copilot" claim by asking what model produces the numbers underneath it. ### FAQ **How can generative AI be used in manufacturing?** Proven 2026 uses: natural-language querying of process and machine data, auto-drafted root-cause and shift-handover reports, alert explanations in operator language, work-instruction generation and upkeep, and maintenance-log summarization. All of them sit on top of structured plant data and industrial ML output. **Can ChatGPT analyze manufacturing process data?** Not reliably as the analysis engine: LLMs are not built to model high-frequency multivariate time-series or discover causal process structure. They excel at explaining and communicating the findings of industrial ML models that are built for that job. **What is the difference between generative AI and industrial machine learning?** Industrial ML learns from a plant's sensor history to detect anomalies and find root causes (numerical, plant-specific, causal. Generative AI learns from text to produce language) explanatory, general, communicative. In a modern factory stack they are complementary layers, not competitors. Before you shortlist Most failed pilots fail on data, not on the model Free, no login. One follow-up from TEEPTRAK, then nothing.Score your data readiness → --- # What Is an Industrial AI Platform? Definition, Categories, ROI (2026 Guide) URL: https://industrialprocessai.com/guides/what-is-industrial-ai/ Summary: A plain-language definition of industrial AI platforms, the four platform categories, what they cost to deploy in time and people, and how to evaluate one in 2026. Translations: fr https://industrialprocessai.com/fr/guides/what-is-industrial-ai/, de https://industrialprocessai.com/de/guides/what-is-industrial-ai/, es https://industrialprocessai.com/es/guides/what-is-industrial-ai/, it https://industrialprocessai.com/it/guides/what-is-industrial-ai/, nl https://industrialprocessai.com/nl/guides/what-is-industrial-ai/, pt https://industrialprocessai.com/pt/guides/what-is-industrial-ai/, pl https://industrialprocessai.com/pl/guides/what-is-industrial-ai/, cs https://industrialprocessai.com/cs/guides/what-is-industrial-ai/, sk https://industrialprocessai.com/sk/guides/what-is-industrial-ai/, hu https://industrialprocessai.com/hu/guides/what-is-industrial-ai/, ro https://industrialprocessai.com/ro/guides/what-is-industrial-ai/, sv https://industrialprocessai.com/sv/guides/what-is-industrial-ai/, tr https://industrialprocessai.com/tr/guides/what-is-industrial-ai/, zh https://industrialprocessai.com/cn/guides/what-is-industrial-ai/ Home / Guides Guide · updated September 1, 2026 ## What is an industrial AI platform? The short answerAn industrial AI platform is software that learns from a factory's own process and machine data to detect anomalies, find root causes and recommend actions: without requiring the plant to employ data scientists. In 2026 the category splits into four types: process-optimization ML, predictive maintenance, machine-data analytics, and enterprise AI platforms. The decisive buying criteria are time-to-value and explainability, not model buzzwords. The TEEPTRAK approach Industrial AI pays where the machine data already exists. The examples on this page share one precondition: a reliable record of what each machine did, minute by minute. TEEPTRAK builds that record without touching the PLC and turns it into loss visibility the plant can act on the same week. - Run, stop and cycle data captured automatically on every machine - Stop reasons that give models labelled history, not only raw signals - Live OEE for operators, so the people who know the machines check data quality every shift Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Score your data readiness → ### The definition that actually helps Strip the buzzwords and an industrial AI platform does three things: it connects to the data a plant already produces (sensors, PLCs, historians, quality systems), it learns what normal looks like for each product and line from that history, and it explains deviations, which variables moved, when, and what to do about it. The explanation step is what separates a platform from an alerting tool. A threshold alarm tells you a temperature is high; an industrial AI platform tells you tonight's scrap spike traces to a raw-material lot interacting with a worn die and a 2°C zone drift. ### The four categories (and which one you actually need) 1 · Process-optimization ML: learns across hundreds of process variables to reduce scrap, stabilize quality and optimize settings. Buy this when your pain is yield, quality drift or unexplained variation. 2 · Predictive maintenance: predicts equipment failure from vibration, electrical and process signals. Buy this when unplanned downtime on critical rotating equipment is the pain. 3 · Machine data & analytics: captures and contextualizes what machines do; the data layer everything else stands on. Example: TEEPTRAK. A module in the electrical cabinet and a sensor at the line, no PLC changes, live OEE in about two hours. 4 · Enterprise AI platforms: build-your-own AI applications on a governed stack. Buy this only with an enterprise program, budget and data team. ### The two numbers that predict success Across the deployments we have reviewed, two variables predict whether an industrial AI project produces ROI: time-to-first-validated-insight (platforms delivering in days keep plant attention; platforms needing quarters lose it) and who operates it daily (tools owned by the plant's own process engineers survive; tools requiring vendor or data-science babysitting decay). Weight your evaluation accordingly: our published rubric puts 40% on exactly these two. ### How to run the evaluation Shortlist three platforms from different philosophies (for example one program-grade incumbent, one fast unsupervised challenger, one category specialist), score them on the six rubric criteria with our Buyer's Scorecard, and insist on a dated first-insight commitment on your line, not a demo dataset. Vendors confident in their time-to-value will accept; the others just told you something important. ### FAQ **How is machine learning used in manufacturing?** The highest-value uses in 2026 are root-cause discovery for scrap and quality drift, predictive maintenance on critical equipment, energy optimization, and automated visual inspection. In each case the model learns the plant's normal behavior from its own history and flags deviations with the contributing variables attached. **Do we need a data scientist to use industrial AI?** Not anymore for the current platform generation. Unsupervised platforms train on the line's own history and are designed for process engineers. Build-your-own enterprise AI stacks still assume data-science staffing. **What does an industrial AI platform cost?** Ranges are wide: machine-data platforms start in the low five figures per year; process-ML platforms typically run mid-five to six figures depending on lines and variables; enterprise AI programs run seven figures. Trial terms vary more than price: the strongest confidence signal is a free or low-risk pilot on your own line. Before you shortlist Most failed pilots fail on data, not on the model Free, no login. One follow-up from TEEPTRAK, then nothing.Score your data readiness → --- # The six criteria for scoring an industrial AI platform URL: https://industrialprocessai.com/methodology/ Summary: The six weighted criteria behind the free Buyer's Scorecard: time-to-value, ML depth, usability, integration, proven outcomes and commercial terms. Translations: fr https://industrialprocessai.com/fr/methodology/, de https://industrialprocessai.com/de/methodology/, es https://industrialprocessai.com/es/methodology/, it https://industrialprocessai.com/it/methodology/, nl https://industrialprocessai.com/nl/methodology/, pt https://industrialprocessai.com/pt/methodology/, pl https://industrialprocessai.com/pl/methodology/, cs https://industrialprocessai.com/cs/methodology/, sk https://industrialprocessai.com/sk/methodology/, hu https://industrialprocessai.com/hu/methodology/, ro https://industrialprocessai.com/ro/methodology/, sv https://industrialprocessai.com/sv/methodology/, tr https://industrialprocessai.com/tr/methodology/, zh https://industrialprocessai.com/cn/methodology/ Methodology ## The six criteria that predict industrial-AI success Use these weights to score any platform in your own demos. They are the same weights built into the Buyer's Scorecard. The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → | Criterion | Weight | What we assess | | Time-to-value | 20% | How fast from installation to the first validated, money-relevant insight. Days beat weeks; quarters fail. | | ML depth & explainability | 20% | Does the platform discover unknown root causes (unsupervised learning across process variables), or only monitor thresholds humans define? Can an engineer see why an alert fired? | | Usability without data scientists | 15% | Can process and reliability engineers operate it alone, or does the vendor quietly assume a dedicated data team? | | Integration & data capture | 15% | Connects to existing sensors, PLCs and historians; hardware requirements; time to first connected line. | | Proven outcomes | 15% | Verified, referenceable results with numbers (scrap %, energy %, downtime hours) across industries and years. | | Commercial accessibility | 15% | Entry price, pilot and trial terms, contract flexibility for mid-market plants, not just enterprises. | ### How to use the rubric Score each vendor 1 to 5 on every criterion, on your own data where possible. Claims without numbers earn nothing on proven outcomes. Ask for a dated commitment to a first validated insight on your line. Re-weight if your plant's priorities differ, but write the new weights down before the first demo. We apply the same criteria when plants evaluate us, starting with time-to-value: live OEE in about two hours, with no PLC changes. Before you scale Twelve questions between a pilot and a line that depends on it Free, no login. One follow-up from TEEPTRAK, then nothing.Get the gate review → --- # OEE system cost: the full 3-year calculator (2026) URL: https://industrialprocessai.com/oee-system-cost/ Summary: What an OEE system really costs over 3 years: licences, devices, sensors, screens, installation, integration and your team’s time. Free calculator. Translations: fr https://industrialprocessai.com/fr/cout-logiciel-trs/, de https://industrialprocessai.com/de/oee-software-kosten/, es https://industrialprocessai.com/es/coste-software-oee/, it https://industrialprocessai.com/it/costo-software-oee/, nl https://industrialprocessai.com/nl/kosten-oee-software/, pt https://industrialprocessai.com/pt/custo-software-oee/, pl https://industrialprocessai.com/pl/ile-kosztuje-system-oee/, cs https://industrialprocessai.com/cs/cena-monitoringu-vyroby/, sk https://industrialprocessai.com/sk/cena-monitoringu-vyroby/, hu https://industrialprocessai.com/hu/oee-meres-koltsege/, ro https://industrialprocessai.com/ro/cost-software-oee/, sv https://industrialprocessai.com/sv/kostnad-oee-system/, tr https://industrialprocessai.com/tr/uretim-takip-programi-fiyatlari/, zh https://industrialprocessai.com/cn/oee-system-cost/ Home / OEE system costFree tool · buyer’s toolkit · no signup for the calculator ## What an OEE system really costs: the 3-year calculator A price per machine is one line of the bill. Devices, sensors, operator screens, installation, integration, training and your own team’s hours decide what you actually pay. Put two offers on the same basis and see the difference before you sign. In shortCompare OEE offers on the total cost of the same scope over the same term, not on the monthly price per machine. A like-for-like figure adds the subscription for every machine, the hardware that makes each machine report, the screens operators use, installation, integration, training, other fees and the hours your own team spends. The calculator below does this for two offers. The Excel model does it for three, shows how the cost moves with plant size and lists 24 questions to send every vendor. ### Compare two offers The example values are illustrative, not any vendor’s price. Replace them with the figures from your quotes and enter every amount in the same currency. #### Your plant Machines to monitorContract term (years)Internal cost of one staff hourCurrency shownEURUSDGBPCHFPLNCZKHUFRONSEKTRYCNY #### Offer A Example: subscription per machine, devices rented, self-installed.Subscription per billed unit per monthBilled units (machines, lines or plants)Machines one device can connectDevice rental per device per monthDevice purchase per device (one-off)New sensors and cabling per machine (one-off)Operator screens or tablets (number)Price per screen or tabletInstallation and on-site work (one-off)Integration with ERP, MES or BI (one-off)Training and onboarding (one-off)Other fees per year (support tier, modules, SSO)Internal hours over the term (set-up, admin, follow-up) #### Offer B Example: lower subscription, devices bought, installation and training included.Subscription per billed unit per monthBilled units (machines, lines or plants)Machines one device can connectDevice rental per device per monthDevice purchase per device (one-off)New sensors and cabling per machine (one-off)Operator screens or tablets (number)Price per screen or tabletInstallation and on-site work (one-off)Integration with ERP, MES or BI (one-off)Training and onboarding (one-off)Other fees per year (support tier, modules, SSO)Internal hours over the term (set-up, admin, follow-up) Offer AOffer B Total over the term·· Year-one cash out·· Equivalent per machine per month·· Share of the total outside the subscription·· · SubscriptionDevicesSensors and screensInstallation, integration, trainingOther feesInternal time Total = one-off costs + term × 12 × (subscription × billed units + device rental × devices) + term × other fees + internal hours × hourly cost. Devices = machines ÷ machines per device, rounded up. Year-one cash out leaves out internal time. ### The cost lines a price list leaves out A published OEE price often covers the software licence only. Ask every vendor to price the same scope line by line, and check who carries each of these costs. | Cost line | Why it matters | Ask every vendor | | Signal pickup and sensors | A machine without a usable signal needs a sensor, cabling or a PLC connection before it can report. | Which of our machines need new sensors or wiring, and who installs them? | | Data devices or gateways | Devices can be rented, bought or included. How many you need depends on how many machines one device connects. | How many devices for our machines, bought or rented, and what happens to them if we stop? | | Operator screens | Stop reasons are entered on a screen at or near the line, and many offers exclude the screens. | Are screens, tablets and mounts included, and which models are supported? | | Network | Devices need Wi-Fi, Ethernet or mobile data on the shop floor. | What network do the devices need, and is a SIM or router included? | | Installation | Self-installation saves money but costs internal hours. On-site work is often quoted separately. | Is installation remote, on-site or done by us, and what does an on-site day cost? | | Integration | Connecting to ERP, MES or BI tools is usually a separate project. | Which integrations are standard, which are quoted, and at what rate? | | Training and language | Operators adopt what they understand in their own language. | In which languages are the software, the training and the support? | | Support | Hours, time zone and channel decide how fast a stopped line gets help. | What are your support hours in our time zone, and in which language? | | Tiers and minimums | Some offers require every licence on the same tier or reserve multi-site features for the top tier. | Can we mix tiers, and which tier do we need for several plants? | | Term and price protection | Multi-year terms lower the monthly price but lock the budget. | What is the price for 1 and 3 years, and how are renewals indexed? | | Exit | Data and hardware must come back cleanly if you change supplier. | Can we export all raw data, and what happens to the devices at the end? | ### Four price structures you will meet Per machine per month Easy to compare, but the total grows with every machine, and devices or sensors are often billed on top. Per line or per plant Flat fees reward monitoring many machines at once. Check what counts as a line and what happens when you add one. Hardware up front, smaller subscription More cash in year one, less every month after. Check the device warranty and who replaces a failed unit. Project quotes for MES-type systems Licence, integration and services priced together, with a wider scope and a longer lead time than a first monitoring step. ### Free trial or paid pilot: what each really costs A free trial is free on the invoice, not in hours: someone still installs, configures, trains operators and reads the data. A paid, supervised pilot on one line costs money but should come with a baseline, agreed success criteria and a decision date. Whichever you choose, write down before you start the result that would make you buy and the result that would make you stop. ### Get the full cost model (Excel) Done: your cost model is downloading. Download the cost model (.xlsx) Next: What one OEE point is worth in your country and Buyer’s Scorecard. #### Get the full cost model (Excel) - Three offers side by side with the same inputs - Cash out year by year and the total over the term - Cost per machine per month at 5, 10, 20, 50 and 100 machines - 24 questions to send every vendor, with a column for each answer ### Questions **How much does OEE software cost?** OEE systems are priced per machine, per line, per plant or per project, so a single monthly figure is not comparable across offers. Ask each vendor for the total cost of the same scope over the same term, with every line item listed, then compare the totals and the cost per machine per month. **What is usually not included in an OEE software price?** Often left out: sensors and cabling, data devices or gateways, operator screens, network equipment, installation, integration with ERP, MES or BI, training, higher support tiers and the time your own team spends. **Is a free trial cheaper than a paid pilot?** Not always. A free trial still uses your team’s hours for installation, configuration and follow-up. A paid, supervised pilot should include a baseline, success criteria and a decision date. Compare both on the hours they need and the decision they let you make. **Should I compare an OEE system with an MES?** Compare them on the scope you need in the next 12 months. An MES covers orders, materials, quality and traceability and is usually priced as a project. An OEE monitoring system covers machine states, stops, speed and counts and is usually the lighter first step. **Which currency should I use in the calculator?** Any, as long as every amount is in the same currency. The calculator does not convert currencies. ### Related tools - Buyer’s Scorecard: score your shortlist on six weighted criteria - RFP question bank - What one OEE point is worth in your country - Grant-ready OEE specification pack - OEE calculator The calculator is vendor-neutral: it applies the same formula to every offer you enter, including TEEPTRAK’s. The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → --- # Pilot-to-Production Gate Review URL: https://industrialprocessai.com/pilot-gate-review/ Summary: Six gates (value, data, integration, people, commercial, risk) each with the evidence that settles it and the owner who has to produce it. Pass or hold, counted for you. Translations: it https://industrialprocessai.com/it/pilot-gate-review/, pt https://industrialprocessai.com/pt/pilot-gate-review/, pl https://industrialprocessai.com/pl/pilot-gate-review/, cs https://industrialprocessai.com/cs/pilot-gate-review/, sk https://industrialprocessai.com/sk/pilot-gate-review/, hu https://industrialprocessai.com/hu/pilot-gate-review/, sv https://industrialprocessai.com/sv/pilot-gate-review/, tr https://industrialprocessai.com/tr/pilot-gate-review/ Free tool ## Twelve questions between a working pilot and a line that depends on it Six gates (value, data, integration, people, commercial, risk) each with the evidence that settles it and the owner who has to produce it. Pass or hold, counted for you. New: What an OEE system really costs: the 3-year calculator Done: your download is ready. Download the Gate Review (.xlsx) Use it after the Buyer's Scorecard: that one picks the vendor, this one decides whether to scale. #### What’s inside - Six gates, twelve decision questions - The evidence that actually settles each one - An owner per question, so nothing floats - Pass/Hold dropdowns with a scale/do-not-scale verdict - Written for the review meeting, not the slide deck The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → --- # Industrial AI RFP Question Bank URL: https://industrialprocessai.com/rfp-questions/ Summary: Grouped by the six rubric criteria we publish, each with what a substantive answer actually contains, so you can tell a real answer from a fluent one. Translations: it https://industrialprocessai.com/it/rfp-questions/, pt https://industrialprocessai.com/pt/rfp-questions/, pl https://industrialprocessai.com/pl/rfp-questions/, cs https://industrialprocessai.com/cs/rfp-questions/, sk https://industrialprocessai.com/sk/rfp-questions/, hu https://industrialprocessai.com/hu/rfp-questions/, sv https://industrialprocessai.com/sv/rfp-questions/, tr https://industrialprocessai.com/tr/rfp-questions/ Free tool ## Nineteen questions vendors cannot answer with a brochure Grouped by the six rubric criteria we publish, each with what a substantive answer actually contains, so you can tell a real answer from a fluent one. New: What an OEE system really costs: the 3-year calculator Done: your download is ready. Download the Question Bank (.xlsx) The criteria and weights come from our published methodology. #### What’s inside - Nineteen questions across six criteria - What a good answer contains, for each - Criterion weights matching the published rubric - A column for the vendor's answer - Includes the exit and data-ownership questions most RFPs omit The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → --- # Industrial AI Platform Buyer's Scorecard (Free xlsx) URL: https://industrialprocessai.com/scorecard/ Summary: Free Excel scorecard: evaluate up to three industrial AI vendors on the six weighted criteria that predict deployment success. Translations: fr https://industrialprocessai.com/fr/scorecard/, de https://industrialprocessai.com/de/scorecard/, es https://industrialprocessai.com/es/scorecard/, it https://industrialprocessai.com/it/scorecard/, nl https://industrialprocessai.com/nl/scorecard/, pt https://industrialprocessai.com/pt/scorecard/, pl https://industrialprocessai.com/pl/scorecard/, cs https://industrialprocessai.com/cs/scorecard/, sk https://industrialprocessai.com/sk/scorecard/, hu https://industrialprocessai.com/hu/scorecard/, ro https://industrialprocessai.com/ro/scorecard/, sv https://industrialprocessai.com/sv/scorecard/, tr https://industrialprocessai.com/tr/scorecard/, zh https://industrialprocessai.com/cn/scorecard/ Free tool ## The Buyer's Scorecard: score your shortlist on six weighted criteria A one-sheet Excel scorecard with the six rubric criteria, fixed weights and a weighted total: built to be filled in during vendor demos and reference calls. New: What an OEE system really costs: the 3-year calculator Done: your scorecard is ready. Download the Scorecard (.xlsx) Pair it with the published rubric. #### What's inside - The six criteria that predict industrial-AI success, with fixed weights (time-to-value and ML explainability at 20% each) - Score three vendors side-by-side, 1-5, with an automatic weighted total - "What good looks like" guidance per criterion: the questions to ask in demos - Red-flag list: the three answers that should end a vendor conversation The TEEPTRAK approach Before you buy an AI platform, get the data layer running. Most shortlists stall on one question: where will the data come from? TEEPTRAK connects machines of any brand and age and delivers states, stops, cycles and OEE from the first days. The AI decision can then rest on measured losses instead of promises. - Works with or without a PLC, on new and old machines alike - Open data export, so any later AI or MES project can use the full history - Operator screens, training and support in the plant's language Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Compare the full cost of offers → --- # Energy per part: what customers ask suppliers for in 2026 URL: https://industrialprocessai.com/we/energy-per-part/ Summary: CDP, Catena-X PCF Rulebook v4.0 and PACT in 2026: the per-part energy and carbon asks, and the measurement chain that makes an answer defensible. Translations: fr https://industrialprocessai.com/fr/we/energie-par-piece/, de https://industrialprocessai.com/de/we/energie-pro-teil/, es https://industrialprocessai.com/es/we/energia-por-pieza/, it https://industrialprocessai.com/it/we/energia-per-pezzo/, nl https://industrialprocessai.com/nl/we/energie-per-stuk/, sv https://industrialprocessai.com/sv/we/energi-per-detalj/, tr https://industrialprocessai.com/tr/we/parca-basina-enerji/ HomeCompliance · Customer carbon and energy data · status 22 September 2026 ## Energy per part: what customers actually ask for in 2026 For plant and operations managers at European suppliers who keep receiving carbon questionnaires. What CDP, Catena-X and PACT ask for, what the CSRD changes mean for a smaller supplier, and the measurement chain that turns an estimate into an answer. In shortThree asks now arrive separately and want the same underlying data. CDP wants organisational energy consumption in MWh with purchased electricity broken down by country. Catena-X wants a cradle-to-gate product carbon footprint per part, with a declared primary-data share and evidence for the electricity claim. PACT wants the same with transparent data reliability metrics. All three are answerable only from metered data you can reconcile to an invoice. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → ### CDP: the questionnaire most suppliers meet first CDP's supply chain programme is the route by which most European suppliers first get asked. More than 200 organisations request data from their suppliers through it, and over 45,000 suppliers were asked to disclose in the 2025 cycle. A bespoke SME questionnaire exists for smaller suppliers, which matters because the full corporate questionnaire is a heavy lift. - The energy question in the full questionnaire sits at Q7.30 and Q7.30.1: report organisational energy consumption in MWh, excluding feedstocks, with purchased or acquired electricity broken down by country or area. - The SME equivalent is Q20.15 and its sub-questions, focused on purchased and acquired electricity. - The 2026 cycle deadline is stated as the week of 26 October 2026, with the exact date and time to be confirmed. - CDP states alignment with ISSB and ESRS, which means the energy figure you produce for sustainability reporting is the same figure CDP wants. One number, produced once, used three times. The practical consequence"Excluding feedstocks", "by country" and "in MWh" are all boundary decisions. If your energy figure is assembled from invoices once a year, the boundary is whatever the invoices happened to cover. If it comes from a metering structure you designed, the boundary is a decision you can defend and repeat. ### Catena-X: the per-part ask, in detail The automotive network's product carbon footprint rulebook, version 4.0, released September 2025, is the most specific demand a European supplier currently faces on per-part data. What it requires, per part: Catena-X PCF Rulebook v4.0: what a supplier has to deliver per part | Requirement | What it means on the shop floor | | Declared unit: for countable products, one piece as described in the part description including a defined weight and the part ID | The footprint is tied to a part number, not to a plant or a product family. Your energy data has to be attributable to a part. | | System boundary: cradle-to-gate, including raw material sourcing, production of materials and components, packaging, disposal of production waste, logistics and internal transport, quality control and manufacturing IT | Internal transport, quality control and manufacturing IT are inside the boundary. Compressed air, lighting and the measuring room are not out of scope. | | Primary data share: report what percentage of the footprint rests on direct measurement rather than industry averages | This is the line that turns machine-level metering into a commercial argument. A supplier with meters reports a high primary-data share; a supplier without them reports averages. | | Allocation hierarchy: avoid allocation, then subdivision, then system expansion with substitution, then physical or economic allocation | Subdivision beats allocation. A meter on the shared compressor is worth more than a formula dividing its consumption by part count. | | Electricity: market-based approach preferred, with a supplier-specific product where a tracking system exists, otherwise the country-specific residual grid mix; renewable commitments need the committed generator's emission factor with supporting documentation; location-based is the fallback, and v4 includes infrastructure emissions | The electricity claim has to be evidenced with contractual instruments or on-site generation records, not asserted. | | Data quality rating covering data age and temporal representativeness, geographic appropriateness, technology and process specificity, and completeness of measurement | Your sampling interval and metering coverage are graded, not just your total. | | Reference period: the time period of primary data collection must be documented | A footprint without a stated collection window cannot be rated. | Source: CX-PCF-Rulebook v4.0, Catena-X Automotive Network, September 2025; standard CX-0029. Paraphrase; the rulebook prevails. Verification is delegated to a separate Catena-X and TfS verification and programme certification framework, which means a footprint can be checked by someone other than the customer who asked for it. ### PACT, and why the two converge PACT, the WBCSD initiative, is the cross-sector equivalent: methodology version 3 and technical specifications for product carbon footprint data exchange version 3.0.3. It uses the same cradle-to-gate boundary, including storage and shipping within and between life-cycle stages, and requires footprints to be accompanied by transparent data reliability metrics giving data quality and the share of primary data. Third-party verification of the calculation model is expected to build trust rather than strictly mandated. Catena-X's rulebook is listed among PACT's resources; the two are deliberately interoperable. What this means for planningYou are not building an answer for one customer scheme. Primary-data share and a documented collection period appear in both. If you meter for one, you have met the substantive requirement of the other, and the remaining work is format. ### The direction of travel: carbon as a technical feature Volkswagen Group's 2025 annual report states that only suppliers who manufacture products in accordance with the requirements of the specifications are eligible for contract award, and that the group plans to make CO2 emissions a technical feature for relevant components, with binding CO2 targets whose compliance must be proved on request. Its S-Rating covers direct suppliers with high sustainability risk and corresponding company size; the 2025 interim target of 85% of procurement volume from relevant suppliers with a positive rating was met at 87%, against a 2040 target above 95% by sales revenue. A separate target has 95% of suppliers holding a certified environmental management system. A technical feature is not a questionnaire. It is a specification value, and specification values are verified on request, held against a tolerance, and treated as nonconformity when missed. That is a different evidentiary standard from an annual sustainability response, and it is the reason to build the measurement chain now rather than when the first specification arrives. ### CSRD after Directive (EU) 2026/470: the cap that lets you say no The Omnibus amending directive, Directive (EU) 2026/470 of 24 February 2026, entered into force on 18 March 2026. For a supplier, one provision matters more than all the rest. The value-chain capProtected undertakings (those with up to 1,000 employees) may refuse information requests that go beyond the standards specified for voluntary use, and the reporting undertaking has to inform them of that right. If a customer sends a smaller supplier a data request built for a large reporting entity, the supplier can decline the parts that exceed the voluntary standard. - Scope thresholds were raised: the net turnover threshold moves from 150 million to 450 million euro, with an employee criterion of more than 1,000 on average during the financial year. We could not confirm from the text we read whether the two criteria apply cumulatively or as alternatives, so verify that before relying on it. - Wave 2 covers financial years starting on or after 1 January 2027, and only for undertakings above the new thresholds. Wave 1 continues with a transition period ending 31 December 2026. - Assurance stays limited. The move to reasonable assurance was removed; the Commission is to adopt assurance standards by 1 July 2027. - Two delegated acts were adopted on 3 July 2026: revised and simplified reporting standards with mandatory datapoints cut by 61%, applying to financial years beginning on or after 1 January 2027, and a separate set of standards for voluntary use by undertakings protected by the value-chain cap. Both were awaiting Official Journal publication when we checked. On the energy disclosure itself: the version operative for financial years 2024 to 2026 asks for total energy consumption in MWh, disaggregated into fossil, nuclear and renewable sources with renewables split three ways, an additional fossil breakdown for high climate impact sectors, separate disclosure of energy production where applicable, and an energy intensity ratio of total energy consumption over net revenue for high-climate-impact activities. We read that paragraph-level detail in a secondary reproduction, not in the official annex, and we could not confirm how the energy disclosure is treated in the revised 2026 standards. Given a 61% datapoint cut, assume the numbering has changed and verify before making claims about reporting for 2027. For the wider picture of what OEMs ask after the Omnibus, see OEM supplier data requests after the Omnibus. Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### The measurement chain that makes an answer defensible Four links, in order. Each one is a place where an answer gets challenged. - Metering granularitySite incomers first, because they reconcile to the invoice. Then every significant energy use individually. Then the machines where energy per part is a decision variable: ovens, furnaces, presses, moulding, compressors, chillers, paint, plating, drying. Shared utilities either get a meter or get a written allocation key, and the Catena-X hierarchy prefers subdivision over allocation, so the meter is worth more. Target: sub-meters explaining at least 90% of the main incomer. - Sampling intervalOne minute or finer at machine level, because anything coarser cannot separate producing from idle and therefore cannot attribute energy to a cycle or a batch. 15 minutes at feeder and area level, matching the European settlement interval so it reconciles with the bill. Monthly at the reporting layer. Keep meter identity, accuracy class and calibration record with the data, and document the collection period: temporal representativeness is graded in the data quality rating. - Allocation to partsEnergy per good part, not per part started. Scrap belongs in the denominator's definition and as a separate variable, so drift is explainable. Attribute by operating state so that idle and changeover energy is visible rather than smeared across the parts that happened to be made. Where one machine makes several part numbers, the allocation key has to be written down and applied consistently, because the declared unit is a part ID with a defined weight. - Reconciliation, machine meter to utility invoiceOne documented chain: machine meters to area meters to site incomers to the utility invoice to the MWh figure reported. Each link with a stated tolerance. This is the single artefact that satisfies the certification auditor, the financial assurance provider under limited assurance, and the customer questionnaire at the same time, and it is the one that is almost never built until someone asks a hard question. The same data, four destinations | What you measure | ISO 50001 | EED audit | CDP / sustainability reporting | Catena-X / PACT | | kWh per good part per machine | Indicator for a significant energy use | Representativeness of the audit | Underlying data for the intensity ratio | Primary data share, per part ID | | Relevant variables measured (volume, mix, state, ambient) | Normalisation | Load profile analysis | Year-on-year comparability | Allocation by subdivision | | Sub-metering to 90% of the incomer | Significant energy uses and the measurement plan | Coverage expectation | MWh split by carrier | Gate-to-gate energy by process | | Meter identity and calibration records | Documented information | Auditor evidence | Assurance | Evidence of electricity source | | Renewable and self-generated split | Energy review | Consumption includes self-consumed renewables | Renewable disaggregation | Market-based versus residual mix | Most plants can start this with the meters they already have plus a handful of clamp meters on the biggest consumers, and a spreadsheet. That is a reasonable first year and it will tell you where the energy actually goes. The point at which it stops working is the reconciliation step, once the number has to be produced monthly, per part number, and defended. - What an OEE system really costs: the 3-year calculator - ISO 50001 evidence: EnPIs, baselines and the measurement plan - OEM supplier data requests after the Omnibus - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - OEM capacity evidence: the standing, portal-reported obligation - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Can a small supplier refuse a customer's carbon questionnaire?** Partly. Directive (EU) 2026/470 gives protected undertakings, those with up to 1,000 employees, the right to refuse information requests that go beyond the standards specified for voluntary use, and requires the reporting undertaking to inform them of that right. It does not release you from a contractual requirement you have already signed, and it does not apply to schemes like Catena-X that a customer imposes through the supply agreement. Read the request against the voluntary standard before answering all of it. **What is a primary data share and why does it matter commercially?** It is the percentage of a product carbon footprint based on direct measurement rather than industry averages, and Catena-X requires it to be reported. It matters because it is a quality score on your answer that the customer can compare across suppliers. A plant with machine-level metering can declare a high share; a plant working from annual invoices and average factors cannot, whatever its actual emissions are. **Does energy per part have to be per good part?** For an energy performance indicator, yes in practice: scrap moves the denominator, so a scrap-blind figure drifts for reasons unrelated to energy performance. For a product carbon footprint, the declared unit is one piece as described in the part description with a defined weight and part ID, and the energy attributed to it has to include the energy spent making the parts that were scrapped, because that energy was really consumed. Keep scrap as an explicit variable so both views are producible from the same data. **How fine does the sampling interval have to be?** One minute or finer at machine level if energy has to be attributed to a part, a cycle or a batch; anything coarser cannot separate producing from idle. 15 minutes at feeder or area level, which matches the European settlement interval and therefore reconciles with the utility bill. Monthly is enough at the reporting layer. The collection period also has to be documented, because temporal representativeness is one of the graded data quality criteria. **What is the single artefact worth building first?** The reconciliation chain: machine meters to area meters to site incomers to the utility invoice to the figure you report, each link with a stated tolerance. It is the only artefact that answers the certification auditor, the assurance provider and the customer questionnaire with the same evidence, and it is the thing that exposes whether your sub-metering actually covers what you think it covers. ### Sources - CDP, Supply Chain programme (200+ requesting organisations; 45,000+ suppliers asked in 2025) - CDP, Disclosure 2026 (final response deadline stated as the week of 26 October 2026) - CDP, Disclosure 2026: Bite-Sized Questionnaire Guidance (Q7.30, Q7.30.1, Q20.15) - Catena-X Automotive Network, CX-PCF-Rulebook v4.0, September 2025 - Catena-X standard CX-0029, Product Carbon Footprint Rulebook - PACT (WBCSD), Methodology and Technical Specifications for PCF Data Exchange v3.0.3 - Directive (EU) 2026/470 of 24 February 2026, in force 18 March 2026 - European Commission, corporate sustainability reporting (delegated acts C(2026) 5010 and C(2026) 5011, adopted 3 July 2026) - Volkswagen Group Annual Report 2025, supplier relationships section Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # OEM Capacity and Audit Evidence Pack (xlsx): free download URL: https://industrialprocessai.com/we/evidence-pack/ Summary: A free xlsx for European suppliers: capacity declaration with demonstrated rate, machine-capability run log, audit evidence index, EnPI and baseline sheet. Translations: fr https://industrialprocessai.com/fr/we/pack-preuves/, de https://industrialprocessai.com/de/we/nachweispaket/, es https://industrialprocessai.com/es/we/pack-de-evidencias/, it https://industrialprocessai.com/it/we/pack-evidenze/, tr https://industrialprocessai.com/tr/we/kanit-paketi/ HomeDownload · OEM capacity and audit evidence · status 22 September 2026 ## OEM Capacity and Audit Evidence Pack One workbook for the records that OEM capacity clauses, VDA 6.3 process audits and ISO 50001 certification each ask for in different words. In shortThe OEM Capacity and Audit Evidence Pack is a free xlsx workbook for European automotive and industrial suppliers. It holds a capacity declaration sheet with demonstrated rate and OEE history, a machine-capability run-conditions log, an audit evidence index mapping requirement to record and location, an EnPI and baseline sheet, and a sources sheet. It organises evidence; it certifies nothing. What is inside the workbook: ### Your file is ready The download starts below. We have also emailed you the link. Download the file - Capacity declaration sheet. One row per line: normal, maximum and current delivery capacity, the demonstrated rate behind the figure, the shift pattern and net available time, shared-asset allocation by part number, and a rolling OEE history with the date of the last demonstration run. - Machine-capability run-conditions log. One row per study: characteristic and classification, machine, tool, material lot, operator, timestamps, sample size, and tick boxes for the conditions claimed (uninterrupted run, single batch, single operator, no tool change) each with a column for where the proving machine record is kept. - Audit evidence index. Requirement, in our own paraphrase, against the record that usually serves as evidence and where yours actually lives. Covers the capacity clauses, VDA 6.3 P6 and corrective-action effectiveness. - EnPI and baseline sheet. Indicator, its significant energy use, the method of determination, the declared baseline period and value, the normalisation variables, and a baseline modification log with a reason and date for every change. - Sources sheet. Every requirement with the document, edition and date it came from, and a column marking what we could not verify. Who it is for. Plant, quality and energy managers at suppliers in Germany, Italy, Spain, France, the Benelux countries and the Nordics who keep capacity, capability and energy evidence in scattered files and want the gaps visible before a customer or an auditor finds them. LimitsIt is not a certification tool and does not replace any standard. IATF 16949, the IATF Rules, VDA 6.3, ISO 22514 and ISO 50001 are copyrighted and sold; check every requirement against your licensed copy and your customer's current portal version. The standards and the customer documents prevail. Nor can a workbook repair the underlying data: if stops are not recorded when they happen, the demonstrated rate will be optimistic. The pack is free. Expect at most one follow-up contact from TEEPTRAK. - What an OEE system really costs: the 3-year calculator - OEM capacity evidence: the standing, portal-reported obligation - Machine capability studies: the conditions, not the maths - VDA 6.3 and production data: where the points go - ISO 50001 evidence: EnPIs, baselines and the measurement plan - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence ### Questions **Is the pack really free?** Yes. It is an xlsx download with no charge and no licence key, and no part of it is gated behind a paid tier. **Will it make us compliant?** No. It organises the evidence you already have and shows where a record is missing. Compliance depends on the standards and the customer documents themselves, which are copyrighted, sold, and prevail over anything in the workbook. **Can we use it if we are not an automotive supplier?** The capacity and capability sheets are written around automotive customer requirements, so parts of them will not apply. The audit evidence index and the EnPI and baseline sheet are general: the ISO 50001 chain is the same in any sector. **Does it need data from a monitoring system?** No. Every sheet can be filled in by hand from production records. The columns that ask where the supporting machine record is kept are there precisely so you can see which claims currently rest on nothing but a number. ### Sources - IATF Global Oversight, Customer-Specific Requirements library - Robert Bosch GmbH, Booklet No. 9: Machine and Process Capability, edition 11.2019 - US DOE / LBNL, 50001 Ready Navigator, Task 11: Energy Performance Indicators and Energy Baselines Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → --- # IATF Rules SI 5: a bad OEM scorecard now costs audit time URL: https://industrialprocessai.com/we/iatf-rules-si5/ Summary: IATF Rules 6th edition SI 5 adds audit hours, or a special audit within 60 days, when OEM scorecard targets are missed. What effective action must show. Translations: fr https://industrialprocessai.com/fr/we/iatf-rules-si5/, de https://industrialprocessai.com/de/we/iatf-rules-si5/, es https://industrialprocessai.com/es/we/reglas-iatf-si-5/, it https://industrialprocessai.com/it/we/regole-iatf-si5/, tr https://industrialprocessai.com/tr/we/iatf-rules-si5/ HomeCompliance · IATF Rules 6th edition · status 22 September 2026 ## IATF Rules 6th edition, SI 5: a bad OEM scorecard is now an audit cost For quality managers and plant directors at IATF 16949 certified suppliers. What SI 5 changed, what the certification body has to do when your OEM scorecard misses target, and what evidence closes it off. In shortUnder the IATF Rules 6th edition, sanctioned interpretation 5, a certification body must add audit time when a certified client misses the IATF OEM quality or delivery targets on the OEM scorecard. If the time cannot be added to the current audit, a special audit must follow within 60 calendar days of the closing meeting. The way out is evidence that systemic corrective action was verified effective. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → What this is based onThe IATF Rules for achieving and maintaining IATF recognition, 6th edition, effective 1 January 2025, are sold, not published. The sanctioned interpretations issued in November 2025 are published on the IATF Global Oversight site. Nothing below is reproduced. The figure of four to eight additional hours comes from a trade-press analysis, not from the Rules text. Check your certification body's own scheme rules; they prevail. ### What SI 5 does The mechanism is simple and it is automatic. Where a certified client fails to meet the IATF OEM quality or delivery targets shown on the OEM scorecard, the certification body must add audit time to investigate the causes and the actions taken. The only escape is evidence that the systemic corrective actions implemented were verified effective. - The trigger is the customer's scorecard, not an internal KPI and not an auditor's impression. The data comes from outside your quality system. - The response is time, not a finding. Reported at four to eight additional hours in trade coverage of the 6th edition. - If the time cannot be added to the audit in progress, a special audit must be held within 60 calendar days of the closing meeting. - The exemption is evidential. Not "we have a corrective action plan" but "the systemic corrective action was implemented and verified effective". Two customer requirements make this concrete. Volvo Group requires a screenshot of its online supplier scorecard to be sent to the IATF certification body ahead of the audit, so the certification body sees the scorecard whether or not you raise it. Stellantis publishes its escalation ladder in its customer-specific requirements: warning letter, escalation level 1, escalation level 2, New Business Hold. IVECO requires its suppliers to notify the certification body within five working days of being placed on Controlled Shipment Level 2 or 3. The wall between commercial performance and certification has been coming down for some time; SI 5 removes what was left of it. ### The other 6th edition changes that pull in the same direction IATF Rules 6th edition: changes that increase the evidence burden | Change | Practical effect | | Extra time allocated in future audits to verify corrective actions for past nonconformities | A nonconformity is not closed when the response is accepted. It is revisited, with the auditor looking for evidence that it worked. | | Corporate multi-site audit-day reduction cut from up to 40% to a maximum of 15% | Group-level systems buy far less relief. Each site carries more of its own audit. | | Audit dates confirmed with the client at least 90 days in advance for surveillance, recertification and transfer audits | More notice, which also means fewer excuses for records that are not retrievable. | | SI 11 and SI 13: final audit reports and all nonconformity management records from the last three years must be available, including evidence that responses were accepted and verification was carried out | Three years of nonconformity history has to be produceable during the audit, not restored from an archive afterwards. | Source: IATF Rules 6th Edition Sanctioned Interpretations, November 2025, and trade-press analysis of the 6th edition, August 2025. ### What "verified effective systemic corrective action" has to look like This is where most suppliers lose the argument, and the certification-body data says so. In the IAOB nonconformance data presented at the 2025 AIAG Quality Summit, the top two IATF 16949 nonconformances were problem solving (10.2.3) and nonconformity and corrective action (10.2.1). One analysis of that data attributes roughly 60% of the 10.2.1 nonconformities to gaps in reviewing effectiveness, and roughly 41% of the 10.2.3 nonconformities to failure to review and update documented information after the fix. The patternNothing in the reported top five is a "we did not do it" finding. They are all "we cannot show that we did it" findings: corrective action plans created but not tracked, missing evidence of implementation, missing evidence of effectiveness verification, incomplete monitoring records. Under SI 5 the evidence has to carry three separate burdens: that the action was systemic, that it was implemented, and that it was effective. Three different records. Three burdens, three records | Burden | What satisfies it | What does not | | Systemic | A root cause stated at process level, with the same cause checked across comparable lines, part families or shifts, and the pFMEA and control plan updated where it applies | A containment action on one machine, and a root cause that names an operator | | Implemented | Dated evidence that the change is live: the revised control plan in use, the parameter limit actually enforced, the new check in the record for every shift since | A signed action plan with a due date in the past | | Effective | A performance series for the affected machine, characteristic or delivery stream spanning the action date, long enough to show the change held | A statement that no further complaints have been received | The third row is the one that needs data you already had to be keeping. You cannot construct a before-and-after series after the fact. If the action was taken in March and the auditor arrives in November, the evidence is the measured series from January to November with the action date marked on it. Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### Why schedule adherence and downtime records became certification-relevant Delivery performance on the OEM scorecard is a logistics number. Under SI 5 it now determines how many hours an auditor spends inside your quality system. That makes the records behind delivery performance part of the certification evidence set, even though they sit in production and logistics rather than quality. - Schedule adherence against the customer requirement, per period, with the cause of each miss. Note that for some customers the delivery date is the arrival date at their goods receiving, not your dispatch date, so the series has to be built on the customer's definition. - Downtime history for the equipment that caused the misses, with stop durations and cause codes captured at the machine. A scorecard miss attributed to "equipment failure" with no underlying stoppage record invites the next question immediately. - The recovery: what the line ran at before, during and after the corrective action, so effectiveness is visible rather than asserted. - Preventive maintenance compliance for the affected assets, planned against done, on schedule. This is the standard maintenance-side evidence request and it is the natural companion to a delivery-miss investigation. Our page on IATF 16949 clause 8.5.1.5 covers the maintenance objectives themselves. None of this is new as a requirement. What SI 5 changes is the cost of not having it to hand. Before, a poor scorecard was a commercial problem managed with the customer. Now it mechanically converts into audit hours spent examining why, and a supplier that cannot answer quickly pays for the answer in time, in a special audit within 60 days, or in a finding. ### What to have ready - Know your scorecard before the auditor doesPull the current OEM scorecard for every certified customer and check it against the target. If a target is missed, assume it will be raised. - Build the miss listEach miss, its date, the cause, the containment, the systemic action and the date it went live. One page per customer. - Attach the seriesFor each systemic action, the measured performance series spanning the action date. This is the effectiveness evidence and the only part that cannot be produced late. - Check the three-year archive is liveSI 11 and SI 13 expect three years of nonconformity management records, including evidence that responses were accepted and verification was carried out, available during the audit. - Use the 90 daysAudit dates are confirmed at least 90 days ahead. That is enough time to close an effectiveness gap with real data, and not enough to invent one. - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - OEM capacity evidence: the standing, portal-reported obligation - VDA 6.3 and production data: where the points go - Machine capability studies: the conditions, not the maths - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Does a missed OEM scorecard target automatically mean a nonconformity?** No. SI 5 requires the certification body to add audit time to investigate causes and actions, not to write a finding. Whether a finding follows depends on what the investigation finds in your system. The cost is certain; the nonconformity is not. **How much extra audit time are we talking about?** Trade coverage of the 6th edition puts it at four to eight additional hours. The Rules themselves are paywalled and we have not quoted them. Ask your certification body what they apply, because they are the ones who will charge it. **What is a special audit within 60 days?** If the additional time cannot be fitted into the audit in progress, a separate audit has to take place within 60 calendar days of the closing meeting. That is a second audit event, with the scheduling and cost that implies, triggered by a customer performance figure. **Our scorecard miss was caused by a customer schedule change. Does that help?** It may explain the cause, but SI 5 is triggered by the scorecard result, not by fault. The useful response is the same either way: the adherence series against the customer requirement, the cause of each miss recorded at the time, and the evidence of what changed afterwards. A documented external cause supported by data is a much stronger position than the same claim made verbally. **Which records should we keep outside the quality system for this?** Schedule adherence measured on the customer's definition of the delivery date, stoppage history with cause codes for the equipment involved, preventive maintenance compliance planned against done for those assets, and a performance series that spans each corrective action date. All four are production and maintenance records rather than quality records, and all four are now part of what an SI 5 investigation looks at. ### Sources - IATF, Rules 6th Edition Sanctioned Interpretations, November 2025 - Quality Magazine, Robert Brown, "The IATF 16949 Rules 6th Edition's Impact on Certification Compliance Auditing", 19 August 2025 - simpleQuE, Paul Gambino, "IATF 16949 Top Nonconformances: Highlights from the 2025 AIAG Quality Summit", 28 October 2025 - Smithers, IATF 16949 news and nonconformance summaries, 2025-2026 - IATF Global Oversight, Customer-Specific Requirements library (Volvo v2, Stellantis v1, IVECO rev. 01) Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # ISO 50001 evidence: EnPIs, baselines and the measurement plan URL: https://industrialprocessai.com/we/iso-50001-evidence/ Summary: What an ISO 50001 auditor asks for: significant energy uses, normalised EnPIs, a declared baseline with a revision log, and the metering plan behind them. Translations: fr https://industrialprocessai.com/fr/we/preuves-iso-50001/, de https://industrialprocessai.com/de/we/iso-50001-nachweise/, es https://industrialprocessai.com/es/we/iso-50001-evidencias/, it https://industrialprocessai.com/it/we/evidenze-iso-50001/, nl https://industrialprocessai.com/nl/we/iso-50001-bewijs/, sv https://industrialprocessai.com/sv/we/iso-50001-bevis/, tr https://industrialprocessai.com/tr/we/iso-50001-kaniti/ HomeCompliance · ISO 50001 and the EED · status 22 September 2026 ## ISO 50001 in evidence terms: EnPIs, the baseline, and the measurement plan For energy managers, plant directors and operations leads at European manufacturers heading into an EnMS certification or an EED energy audit. What the evidence chain has to contain, why absolute kWh will not do, and the deadlines that are forcing the question. In shortAn ISO 50001 audit follows one chain: the energy review identifies significant energy uses, each significant use gets an energy performance indicator, each indicator has a declared baseline period, and a measurement plan says what is metered, where, how often and how accurately. Where relevant variables affect consumption, the indicator must be normalised. Absolute kWh on its own is not evidence of performance. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → What this is based onISO 50001:2018, ISO 50006:2023 and ISO 50015 are copyrighted and sold. Nothing here reproduces standard text. The clause structure comes from the official ISO contents listings; the substance on indicators, baselines and normalisation comes from the US Department of Energy's 50001 Ready Navigator, which is official implementation guidance for the standard. The standards prevail. Buy them before building a system on this page. ### The evidence chain, in order An EnMS audit is not a search for savings. It is a check that a chain holds together. Each link produces documented information, and a break in one link makes the links after it unverifiable. ISO 50001:2018 clauses that carry the evidence | Clause | What it does | What has to exist | | 6.3 Energy review | Systematic analysis of energy use and consumption; identifies significant energy uses and improvement opportunities | The analysis itself, the data it used, and the resulting list of significant energy uses with the criteria that put them on the list | | 6.4 Energy performance indicators | The metrics used to monitor performance and demonstrate improvement | The indicator list, the method for determining and updating each one, and the periodic values | | 6.5 Energy baseline | The quantitative reference each indicator is compared against | The baseline period, the baseline value, the conditions under which it will be revised, and records of every revision | | 6.6 Planning for collection of energy data | The measurement plan | What is metered, where, at what interval, with what accuracy, and how the data is stored | | 7.5 Documented information | Creation, updating and control | Meter identity, calibration status, and the version history of the method documents | | 9.1 Monitoring, measurement, analysis and evaluation | Evaluation of energy performance and of the EnMS | Indicator values against baseline over time, with the analysis and the actions taken | Clause structure from the official ISO 50001:2018 contents listing. Paraphrase; the standard prevails. ### Why absolute kWh is not acceptable where relevant variables exist An indicator can be a single metric, a ratio, or a numerical model. Official implementation guidance says each significant energy use should have an associated indicator, and that where relevant variables significantly affect consumption, the indicator must be normalised. Absolute consumption on its own is not sufficient evidence of performance. The reason is mechanical. Site electricity fell 8% last year. Did energy performance improve? Not if output fell 11%. Not if the product mix moved to a lighter part. Not if last winter was mild and this one was milder. A plant that reports absolute kWh is reporting the combined effect of performance and everything else, and cannot separate them when asked. What normalisation means in practiceIt means dividing or modelling out the things that move consumption for reasons other than how well you run. Accepted methods include energy intensity ratios, regression, or another documented adjustment. Where a regression model is used, keep the model, its coefficients, its R-squared and its residual behaviour: model uncertainty is explicitly part of the subject, and it is usually the first thing a competent certification auditor asks about after seeing the indicator list. The practical minimum set of relevant variables for a discrete manufacturer: - Production volume in good parts, or kilograms of good output, per time bucket. Per good part, not per part started: scrap moves the denominator, and a scrap-blind indicator drifts for reasons that have nothing to do with energy. - Product mix or part number. Energy per part is not comparable across part families, so a single plant-level figure hides mix effects entirely. - Operating state duration: producing, idle, changeover, planned down, unplanned down. Idle share is usually the largest recoverable loss and it is invisible without state data. - Ambient temperature, and humidity where compressed air or HVAC is a significant energy use. - Shift calendar and operating hours. - For process plants, add throughput, batch size, feedstock quality and set-point temperatures. ### The baseline, and the modification log nobody keeps The baseline is the indicator's value over a declared baseline period. Three obligations follow, and the third is the one most often missing at certification. - Declare the period. Twelve months is the defensible default for a plant with seasonal load and an annual product-mix cycle. Anything shorter has to be justified against the relevant variables. - Write the revision conditions in advance. Guidance calls for a predetermined revision methodology, not a decision made when the numbers become inconvenient. - Keep records of every baseline modification. This is an explicit documentation requirement, and it is what allows an auditor to trust a year-over-year comparison. Baselines must be revised when the indicators no longer reflect actual performance, or when there are major changes to static factors, processes, operating patterns or energy systems. Static factor is the term that matters on a shop floor: a new line installed, a shift pattern changed, a building envelope changed, a product transferred in or out. Each of those is a legitimate reason to rebaseline, and each of them needs an entry in the log saying what changed, when, what the old and new baseline values were, and on what method. The testA rebaseline with a written reason and a dated entry is good practice. A rebaseline with no entry looks like moving the goalposts, and it is indistinguishable from moving the goalposts in the records. ### The measurement plan: what has to be metered Clause 6.6 is the clause that forces the organisation to decide what is metered, where, how often and with what accuracy. A practical hierarchy, working down: - Site main incomers: electricity, gas, district heat or steam, compressed air generation, water. These reconcile to the invoices, and invoiced energy is what the EED counts. - Each significant energy use identified in the energy review, metered individually. If a use is significant enough to be on the list, it is significant enough to measure. - Each machine or cell that is both a large consumer and a place where energy per part is a decision variable: ovens, furnaces, presses, injection moulding, compressors, chillers, paint, plating, drying. - Shared utilities that get allocated: compressed air, chilled water, steam. Either sub-meter them or document a defensible allocation key. A meter is worth more than a formula, and it is worth more in customer carbon reporting too. A workable target is enough sub-metering that the sum of sub-meters explains at least 90% of the main incomer. That clears the German draft's proposed 90% audit coverage and the EU guidance expectation comfortably, and it is the point at which the number you report stops being an estimate with a meter attached. Sampling interval by layer, and what each interval is for | Layer | Interval | Why | | Machine or cell | One minute or finer | Anything coarser cannot separate producing from idle, and cannot attribute energy to a cycle, a part or a batch | | Feeder or area | 15 minutes | Matches the European settlement interval, so it reconciles with the utility bill and supports load-profile analysis | | Reporting layer | Monthly | Indicator against baseline, and the annual roll-up for corporate and customer reporting | Retention: at least the baseline period plus the full audit cycle. With audits every four years and a twelve-month baseline that is a minimum of five years of interval data, and realistically the life of the baseline. Keep meter identity, location, accuracy class and calibration record with the data. An auditor who cannot tell which physical meter produced a number will not accept the number. Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### The EED duties and dates that force the issue Directive (EU) 2023/1791 sets two consumption thresholds. Member States had to notify transposition by 11 October 2025 and several have not, so the rule that applies to your plant is national, not European. Consumption is measured as the average annual final energy consumption over the previous three years, and counts energy invoiced to the enterprise plus self-consumption of self-produced renewable energy. EED thresholds and the dates behind them | Trigger | Duty | Deadline | | Above 85 TJ average annual final energy consumption | Energy management system certified by an independent body; ISO 50001 is the standard the Commission guidance points to | 11 October 2027 | | Above 10 TJ and no EnMS | Energy audit, repeated at least every four years | First audit 11 October 2026 | | Newly obliged enterprises crossing a threshold later | Two years to put the EnMS in place, one year to complete the audit | From the date of becoming obliged | Commission Recommendation (EU) 2024/2002 of 24 July 2024, the Commission's own interpretation guidance on Article 11. 85 TJ is 23.61 GWh; 10 TJ is 2.78 GWh. - Denmark has the sharpest date in Europe. Under Bekendtgørelse nr. 1138 of 18 September 2025, in force 11 October 2025, a certified energy management system is required above 85 TJ by 1 August 2026: fourteen months ahead of the EU date. Audits above 10 TJ were due by 1 August 2025, then every four years. A separate climate audit applies above 570 tonnes of CO2 a year, also from 1 August 2026, and audit reports, action plans and key performance indicators are filed through virk.dk. - Germany's reform is in progress and is not law. As of 22 September 2026 the EnEfG obligation still bites at 7.5 GWh a year on a three-calendar-year average. The bill that would move it to 23.6 GWh and make the audit duty consumption-based at about 2.77 GWh with at least 90% coverage was introduced as Bundestag Drucksache 21/8027 on 16 September 2026 and was scheduled for first reading on 24 September 2026. Until it is promulgated, 7.5 GWh applies. - Finland is already in force. Under the energy efficiency act 1372/2025, in force 1 January 2026, an audit is required above 2,700 MWh a year with a publicly available action plan, and a certified management system above 23,600 MWh. First audits for newly qualifying companies 11 October 2026, EnMS 11 October 2027. - Several countries have not transposed. Italy, the Netherlands, Austria, Spain and Wallonia were all still on older or draft regimes when we checked on 22 September 2026, and Sweden's agency page still pointed to the 2014 law. Check the national rule, not the directive. The audit side has its own coverage expectation: Commission guidance describes covering either 65 to 90% of total energy consumption, or all consumption areas above 10% each totalling at least 80%. The German draft proposes at least 90%. That is the same argument as the metering hierarchy above, arriving from a different direction. ### The three gaps that show up at certification - Indicators that exist but have no method document. The method for determining and updating each indicator is itself required documented information. A number in a monthly pack with no written definition is not an indicator, it is a figure. - A baseline that has quietly moved. No modification log, so no year-over-year comparison can be relied on. - Data that cannot be tied to a meter. A spreadsheet column labelled "compressor" with no meter identity, accuracy class or calibration record behind it. The reading may be perfectly correct and it is still not evidence. None of those three are solved by buying meters. They are solved by writing down what the indicator means, when the baseline changes, and which meter produced which column, and then keeping the raw series long enough that someone can check. - What an OEE system really costs: the 3-year calculator - Energy per part: what customers ask for in 2026 - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - OEM supplier data requests after the Omnibus - OEM capacity evidence: the standing, portal-reported obligation - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Does ISO 50001 require energy per part?** It requires energy performance indicators, normalised where relevant variables significantly affect consumption, and official implementation guidance says each significant energy use should have one. For a discrete manufacturer whose significant energy uses are production machines, energy per good part is the natural indicator, but it is not named in the standard. What is required is that the indicator you choose actually isolates performance from output, mix and weather. **Can we use absolute kWh as an EnPI?** Only where no relevant variable significantly affects consumption. That is rare in manufacturing. Where output, product mix, operating hours or ambient conditions move your consumption, the indicator has to be normalised, and absolute consumption on its own is not sufficient evidence of energy performance. **When do we have to change the baseline?** When the indicators no longer reflect actual performance, when static factors change significantly (a new line, a changed shift pattern, a changed building envelope, a product transferred in or out) or according to the revision method you wrote down in advance. Every modification needs a retained record. The log matters as much as the revision. **Which EED deadline applies to our plant?** The national one. The directive sets certified EnMS above 85 TJ by 11 October 2027 and a first audit above 10 TJ by 11 October 2026, but Member States implement it differently and several were late. Denmark requires the certified system by 1 August 2026. Germany was still applying its existing 7.5 GWh trigger on 22 September 2026 while a reform bill was at first reading. Finland is in force from 1 January 2026. Check your national rule before you plan around a European date. **How much sub-metering is enough?** Enough that the sum of your sub-meters explains at least 90% of the main incomer. That satisfies the German draft's proposed 90% audit coverage and the EU guidance expectation of roughly 80% or more, and it is the point at which the allocation to individual machines stops being an estimate. Anything shared and allocated (compressed air, chilled water, steam) either gets a meter or gets a written, defensible allocation key. ### Sources - ISO 50001:2018, Energy management systems: Requirements with guidance for use - ISO 50006:2023, Evaluating energy performance using energy performance indicators and energy baselines - ISO 50015:2014, Measurement and verification of energy performance of organizations - US DOE / LBNL, 50001 Ready Navigator, Task 11: Energy Performance Indicators and Energy Baselines - Commission Recommendation (EU) 2024/2002 of 24 July 2024 on Article 11 of Directive (EU) 2023/1791 - European Commission, Energy Efficiency Directive overview (transposition deadline 11 October 2025) - Denmark, Bekendtgørelse nr. 1138 of 18 September 2025, in force 11 October 2025 - Germany, Bundestag Drucksache 21/8027, 16 September 2026 (EnEfG/EDL-G reform bill, first reading) - BAFA, Energieaudit (EnEfG and EDL-G duties in force), read 22 September 2026 - Energiavirasto, new energy efficiency obligations under act 1372/2025, news item 3 June 2026 Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # Machine capability study: the conditions that make Cmk hold URL: https://industrialprocessai.com/we/machine-capability-conditions/ Summary: Cm/Cmk on 50-100 parts from one uninterrupted run, Cp/Cpk and Pp/Ppk over 20 days, Mercedes DS/DZ values, and the run records that make the index defensible. Translations: fr https://industrialprocessai.com/fr/we/capabilite-machine-conditions/, de https://industrialprocessai.com/de/we/maschinenfaehigkeit-bedingungen/, es https://industrialprocessai.com/es/we/capacidad-de-maquina-condiciones/, it https://industrialprocessai.com/it/we/capability-macchina-condizioni/, tr https://industrialprocessai.com/tr/we/makine-yeterliligi-kosullari/ HomeCompliance · Machine and process capability · status 22 September 2026 ## The machine capability study is a conditions problem, not a maths problem For quality engineers and plant managers at automotive and industrial suppliers. What the published acceptance values are, whose rules they come from, and what records you have to be able to produce to show the study ran under the conditions it claims. In shortAnyone can compute Cmk. The index means something only if the run behind it was uninterrupted, on one material batch, with one operator and no tool change, over 50 to 100 consecutive parts. Bosch Booklet No. 9 sets Cm and Cmk at 1.67 and the follow-on Cp/Cpk and Pp/Ppk at 1.33. The evidence that decides the study is the machine record of the run window. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → What this is based onBosch Booklet No. 9, "Machine and Process Capability", edition 11.2019, is published openly and is the clearest real rule set available. Customer requirements from Mercedes-Benz, Volvo Group, Stellantis and Ficosa are read from their own published documents. Nothing is reproduced, only paraphrased. ISO 22514, the customer document and your licensed standard prevail over this page. ### The acceptance values, and where each one comes from There is no single European number. There is a widely used default, and then customer-specific values that override it. Both matter, because a supplier that applies the default to a characteristic the customer has classified differently has done the study for nothing. Published capability acceptance values and their sample rules | Study | Sample rule | Acceptance | Source | | Machine capability, Cm and Cmk | At least 50 parts, 100 recommended, from one continuous run under conditions that suppress non-machine influences | Cm and Cmk at least 1.67 | Bosch Booklet No. 9, edition 11.2019 | | Long-term process capability, Cp and Cpk | At least 125 measurements, typically 25 samples of 5, over a sufficiently long period; roughly 20 production days as a guide | Cp and Cpk at least 1.33 | Bosch Booklet No. 9, edition 11.2019 | | Long-term process performance, Pp and Ppk | Same 125-measurement rule | Pp and Ppk at least 1.33 | Bosch Booklet No. 9, edition 11.2019 | | Measurable DS/DZ characteristics | In deviation from the industry standard | Ppk or Pmk at least 2.0; for stable processes Cpk at least 1.67; 100% inspection until capability is verified | Mercedes-Benz Special Terms, MBST 13/19 section 5, edition 2019-11 | | Special characteristics | At least 30 pieces from the significant production run | Cpk study required | Volvo Group CSR v2, clause 8.3.4.4, 28 February 2025 | | Special-characteristic ("CO") parts | At least 125 units, at least 2 per independent cavity or tool | Ppk above 1.67 at initial approval, Cpk above 1.33 thereafter | Ficosa Supplier Quality Manual, edition 6, 20 February 2026 | | Characteristics on the Product/Process Information Sheet | Inspection frequencies aligned to the targets | Cp and Cpk targets carried on the sheet itself | Stellantis CSR v1, clause 8.2.3.1.2, June 2025 | Ficosa is a Tier 1 setting requirements on its own suppliers, not an OEM. It is here because it shows how OEM requirements cascade, not as evidence of what an OEM demands. Two things follow. First, the Mercedes rule has an operational consequence, not just a numerical one: until capability is verified for a measurable DS or DZ characteristic, those characteristics are inspected 100%. That makes the verification date itself a record, and every part made before it needs the 100% inspection record behind it. Second, the Volvo and Ficosa rules define which parts, not just how many: parts from the identified significant production run, and at least two per cavity. A sample of the right size drawn from the wrong population fails. ### What "uninterrupted, single batch, single operator, no tool change" means in records The machine study deliberately excludes the disturbances that series production will add. That is the whole point of it: it isolates the machine. Every source that describes the study specifies the conditions, and the conditions are what you have to be able to evidence. Study condition to the record that proves it | Condition | What has to be shown | How it usually fails | | Uninterrupted run | A machine timeline for the study window with no stoppage events between the first and last part | A short stop nobody logged. The measurement file is continuous because measurements are continuous; the run was not. | | Consecutive parts | Part sequence tied to the run, in production order | Parts measured later in whatever order they came off the tray, so sequence information is lost and any trend is unreadable | | One material batch | The material lot consumed during the window, from the material record | Two lots crossed mid-run and nobody recorded the change | | One operator | Who was logged on for the window | Shift change inside the window | | No tool change | Tool or insert change events for the asset, showing none in the window | A worn insert swapped during the run because it was the sensible thing to do at the time | | Stated time window | Start and end timestamps for the run, matching the measurement records | A study dated to the day, not the window, so it cannot be checked against anything | The practical pointA spreadsheet of 50 measurements proves that 50 parts were measured. It does not prove that the run was uninterrupted, that the batch did not change or that no tool was swapped. A continuous machine record for the same window does. If the conditions cannot be evidenced, the index is a number without a claim attached. ### The obligation does not stop at release Bosch requires the process to remain permanently stable after release, monitored with control charts, with the reactions defined in the control plan and documented when they happen. In audit terms that produces three continuing records: - The control-chart record itself, retained, with every rule violation visible rather than smoothed out of a monthly summary. - The reaction to each violation: what was stopped, when, and what happened to the parts made since the previous good check. Auditors reviewing SPC charts and capability values expect a reaction plan where capability falls below requirement, and expect to see it used. - The actual check frequency against the planned frequency, per characteristic. Stellantis aligns inspection frequency to the Cp and Cpk targets on the Product/Process Information Sheet, which makes the achieved frequency an auditable number in its own right. The recurring weakness here is not the statistics. It is that the measurement record and the production record live in different systems, so nobody can say which machine made the part that produced a given reading, in which time window. Where a capability value cannot be resolved to a characteristic, an operation, a machine and a period, it cannot be defended. For the maintenance side of the same argument, see our page on IATF 16949 clause 8.5.1.5 and TPM objectives. Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### Correction: VDA 5 is not about machine capability This mis-citation circulates widely, including in vendor articles that otherwise get the numbers right. VDA Volume 5 is about measurement and inspection process suitability (Prüfprozesseignung) currently in its 3rd edition, July 2021, with a practical guide published in September 2022. It tells you whether your measuring equipment and measurement process are good enough to judge the characteristic. It does not set Cmk limits. The documents that do sit behind machine and process capability are ISO 22514, specifically ISO 22514-3 for machine performance studies on measured data from discrete parts, and VDA Volume 4 for VDA's process-assurance methods. Bosch Booklet No. 9 cites ISO 22514-2 as its primary standard alongside ISO 22514-1, -3, -5 and -6, ISO 9001, VDA 4 and the AIAG SPC and PPAP manuals. If someone hands you a capability requirement attributed to VDA 5, ask which document it actually came from before you design a study around it. The distinction is not pedantic. A measurement process that is not suitable produces capability indices that are wrong in an unknown direction, and measurement systems analysis is consistently among the most frequent IATF 16949 nonconformances. Getting VDA 5 right is a prerequisite for the capability study, not a substitute for it. ### The terminology change to watch, and its current status A joint AIAG and VDA SPC manual is in draft as a yellow volume for public review, with a feedback window that ran to 3 May 2026. Its stated intent is to reconcile the long-standing AIAG and VDA differences and to align with ISO 22514, ISO 3534 and ISO 7870. - It replaces Cm and Cmk with Pm and Pmk, on the reasoning that a short-term machine trial says nothing about stability, so the result should be reported as performance rather than capability. - It sets out a sequence: machine performance at qualification, preliminary process performance at qualification, then continuous monitoring with Cp/Cpk and Pp/Ppk after launch. - It moves towards control charts with adjustable significance levels rather than fixed limits that depend on sample size. Status, plainlyThis is a draft in review, not a published requirement. We did not confirm whether a final version has issued since the May 2026 feedback deadline. Do not rewrite your capability procedure on the strength of it, and do not let a supplier or a customer tell you Cmk has been abolished. Check the published volume before changing anything. ### What to have ready before the study - Confirm the characteristic classification with the customerWhether a characteristic is DS, DZ, special or ordinary decides which acceptance value applies. Mercedes at 2.0 and a default at 1.67 are different studies. - Confirm the measurement process is suitable firstVDA 5 or the equivalent MSA work comes before the capability study, not after it. - Fix and record the run windowStart and end timestamps, machine, tool, material lot, operator, parameter set. Write them down before the run, not from memory afterwards. - Keep the parts in sequenceLabel in production order. Without sequence there is no trend analysis and no way to relate a drift to anything that happened on the machine. - Pull the machine record for the same windowStoppages, tool changes, operator logons, parameter changes. This is the document that turns the index into evidence. - Plan the follow-on study nowThe 125-measurement process study over roughly 20 production days is a separate exercise with a separate sampling scheme. Decide the scheme before series production starts, not when the customer asks. - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - OEM capacity evidence: the standing, portal-reported obligation - VDA 6.3 and production data: where the points go - IATF Rules 6th edition SI 5: a bad scorecard is now audit time - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Is Cmk 1.67 a rule or a convention?** It is a published requirement in Bosch Booklet No. 9, edition 11.2019, which is a real Tier-1 rule set applied to Bosch's suppliers, and it is widely used as a default across German automotive. It is not a universal legal threshold. Your customer's own document decides, and some customers are stricter: Mercedes-Benz requires Ppk or Pmk of at least 2.0 for measurable DS and DZ characteristics. **How many parts does a machine capability study need?** Bosch specifies at least 50 parts, with 100 recommended, from one continuous run. Volvo requires at least 30 pieces for Cpk studies on special characteristics taken from the significant production run. Ficosa requires at least 125 units with at least two per independent cavity or tool. The sample rule travels with the customer, and so does the population the sample must be drawn from. **Does VDA 5 set Cmk limits?** No. VDA Volume 5 covers measurement and inspection process suitability, currently in its 3rd edition of July 2021. The attribution of Cmk limit values to VDA 5 appears in at least one widely read vendor article and is wrong. Machine and process capability sit with ISO 22514 and VDA Volume 4, and the openly published acceptance values used across German automotive come from Bosch Booklet No. 9. **Has Cmk been replaced by Pmk?** Not as of 22 September 2026, as far as we could verify. A joint AIAG and VDA SPC manual is in draft as a yellow volume, with a public review window that closed on 3 May 2026, and that draft does replace Cm and Cmk with Pm and Pmk. We did not confirm whether a final version has been published. Until it is, the existing customer requirement applies. **What do we have to keep if the customer never asks for the raw run?** Keep the run window record anyway: timestamps, machine, tool, material lot, operator, parameter set, and the machine's own stoppage and tool-change record for the same window. Most customers ask for the index. The ones who challenge it ask how the study was run, and that question cannot be answered retrospectively. ### Sources - Robert Bosch GmbH, Booklet No. 9: Machine and Process Capability, edition 11.2019 - ISO 22514-3, Machine performance studies for measured data on discrete parts - Mercedes-Benz Special Terms 2020, MBST 13/19, document ALD00001238, edition 2019-11 - Volvo Group, Customer Specific Requirements v2, 28 February 2025 (IATF CSR library) - Stellantis, CSR for use with IATF 16949, v1, June 2025 - Ficosa, Supplier Quality Manual, I-CP-XX/XX-16 edition 6, 20 February 2026 - VDA QMC publications (VDA 5, 3rd edition July 2021; VDA 6.3, 4th edition January 2023) - Zbigniew Huber, superengineer.net: AIAG-VDA SPC yellow volume draft for review, February 2026 Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # OEM capacity evidence: what suppliers must prove in 2026 URL: https://industrialprocessai.com/we/oem-capacity-evidence/ Summary: Mercedes SMB/BKM, IVECO clause 17, Volvo plant capacity assessment and Stellantis escalation: what evidence backs a capacity figure, and what fails. Translations: fr https://industrialprocessai.com/fr/we/preuves-capacite-oem/, de https://industrialprocessai.com/de/we/oem-kapazitaetsnachweis/, es https://industrialprocessai.com/es/we/evidencia-de-capacidad-oem/, it https://industrialprocessai.com/it/we/evidenze-capacita-oem/, pl https://industrialprocessai.com/pl/we/zdolnosc-produkcyjna-oem/, ro https://industrialprocessai.com/ro/we/dovezi-capacitate-oem/, tr https://industrialprocessai.com/tr/we/oem-kapasite-kaniti/ HomeCompliance · OEM capacity requirements · status 22 September 2026 ## OEM capacity evidence: capacity is a standing, portal-reported obligation For plant managers, logistics managers and supplier-quality leads at European automotive suppliers. What four openly published OEM requirement sets actually ask for on capacity, what record supports a capacity claim, and where a spreadsheet figure breaks down. In shortCapacity is not a launch document. Mercedes-Benz requires normal, maximum and current delivery capacity in its SMB portal, kept promptly, plausibly and completely. IVECO requires evidence of maintaining required capacity on dedicated and shared lines. Volvo requires a plant capacity assessment before an RFP is answered. What satisfies all three is a demonstrated rate measured on the line, not a figure rebuilt afterwards in a spreadsheet. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → What this is based onThe OEM documents named here are published openly, most of them through the IATF Global Oversight customer-specific requirements library. They are copyrighted; nothing below is reproduced, only paraphrased, and the current version on the customer's own portal prevails. Where a requirement sits behind a supplier login and could not be read, this page says so rather than guessing. ### Capacity is a standing obligation, not a launch document Most suppliers treat capacity as a PPAP artefact: a run-at-rate at launch, a signed capacity sheet, then silence until the next change. Three of the openly published European customer-specific requirement sets do not read that way. They ask for capacity to be maintained, reported and re-proved through serial life, and they name the portal it goes into. Capacity duties in openly published European OEM requirement documents (versions listed on the IATF library, 22 September 2026) | Customer | Document | What it asks for on capacity | Where it lands | | Mercedes-Benz | Mercedes-Benz Special Terms, document ALD00001238, edition 2019-11 | Normal capacity, maximum capacity and current delivery capacity entered for specified periods; changes mapped promptly, plausibly and completely; shortfalls communicated with concrete alternatives | SMB supplier portal, module BKM (demand and capacity management); shortfalls in module BBM (demand and inventory management) | | IVECO Group | CSR for IATF 16949, revision 01, 24 November 2025 | Clause 17, against IATF 7.1.3.1: evidence of maintaining the productive capacity IVECO requires, for dedicated and for non-dedicated (shared) production lines | Supplier evidence on request; performance indicators tracked in IVECO SQP | | Volvo Group | CSR, version 2, 28 February 2025 | Clause 8.2.3.1.3: manufacturing feasibility before responding to any RFP, including a plant capacity assessment | Feasibility record held by the supplier; scorecard screenshot sent to the IATF certification body before the audit | | Stellantis | CSR for use with IATF 16949, version 1, June 2025 (effective July 2025) | The CSR carries Cp and Cpk targets on the Product/Process Information Sheet and a named escalation ladder; the capacity method itself sits in the Capacity Assessment Tool, outside the CSR | GST, Amadeus/SQP and the other named Stellantis systems | Source: IATF Global Oversight CSR library and the Mercedes-Benz DocMaster public portal, both read 22 September 2026. Read the four rows together and the pattern is one obligation in four shapes: the customer wants a capacity number that is current, that it can compare with what you are actually shipping, and that you can explain. ### Mercedes-Benz: the figure lives in the portal, and it has to be plausible The Mercedes-Benz IATF customer-specific requirements document is thin and mostly points elsewhere. The substance for capacity is in the openly published Mercedes-Benz Special Terms. What that document asks of a production partner: - Three capacity figures, not one. Normal capacity, maximum capacity and current delivery capacity, for the periods specified, entered into the SMB portal module BKM. - Changes mapped promptly, plausibly and completely. Those three words are the requirement. Promptly is a timing test, completely is a coverage test, and plausibly is an evidence test. - Shortfalls go into BBM with alternatives. If quantities or dates cannot be met, the partner communicates that in the demand and inventory management module together with concrete alternatives. A warning without an alternative is not the deliverable. - Delivery days are arrival dates at Mercedes goods receiving, not dispatch dates from your yard. Schedule adherence is measured on the customer's clock. - Records retained per VDA Volume 1 retention periods. The word to plan around"Plausibly" is the one that decides what you need to keep. A capacity figure is plausible when the availability, the rate and the quality yield that produce it can be shown for the same line and the same period. It stops being plausible the moment the declared number and the shipped quantity move in different directions with no recorded reason. One caveat on editions. The DocMaster public portal also lists a Mercedes-Benz Special Terms 2025 edition (document ALD00002454, edition 2025-04) that we did not read. The clauses above come from the 2019-11 edition of ALD00001238. Check the edition that applies to your contract before you build a process on this. ### IVECO, Volvo, Stellantis: three more shapes of the same duty - IVECO clause 17 covers shared lines explicitly. If a press makes parts for three customers, the capacity you declare to IVECO is what is left after the other two, and you have to show the split. Clause 15 adds a separate duty: notify your certification body within five working days of being placed on Controlled Shipment Level 2 or Level 3. - Volvo puts the capacity assessment before the quotation. Manufacturing feasibility including a plant capacity assessment is required before answering an RFP, which reverses the usual order. Volvo also runs a three-stage Low Performing Supplier process triggered by negative performance trends, and requires a screenshot of the online supplier scorecard to go to the IATF certification body ahead of the audit. - Stellantis names the ladder. Warning letter, escalation level 1, escalation level 2, New Business Hold, written into the CSR at 9.1.2.1, so the consequence of missing schedule is defined in advance. Stellantis also requires dimensional reports as structured, digitised data, which is an explicit signal that a scan is not the deliverable. The Stellantis capacity method itself, the Capacity Assessment Tool, is not in the CSR. What we found describing it is a training provider's undated course description, which puts CAT inside the APQP and PPAP flow and has it re-run during run-at-rate, during development and in serial life whenever process, equipment or location changes. Treat that as secondary. Ask your Stellantis supplier-quality contact for the current CAT instruction. ### What evidence satisfies a capacity claim The clearest open specification of the method is a Tier-1 manual rather than an OEM one: the Plastic Omnium Lighting Capacity Analysis Report manual, document LSM-SCM-403 revision 05, 8 November 2022. Its pass condition is a single comparison. The pass conditionDemonstrated OEE must be at least the required OEE. Required OEE is derived backwards from customer demand, grossed up for downstream scrap, against the maximum parts theoretically possible in the net available time. Demonstrated OEE is availability multiplied by performance efficiency multiplied by quality rate, measured on the run. For the definitions themselves, see manufacturing-metrics.org. That comparison only works if you have real numbers for the inputs. The CAR manual lists them: net ideal cycle time in seconds per part, scrap rate per process, in-line rework rate, changeover time and frequency, net available time after planned downtime, observed unplanned downtime, and the shift and day pattern. It also asks for a historical manufacturing performance sheet carrying 25 data points representing 25 weeks of production, to establish an average-OEE baseline, and a shared loading plan for shared equipment, with total allocation above 90% treated as a yellow flag. Capacity claim to supporting record | The claim you make | The record that supports it | What gets challenged | | "This line runs at X parts per hour" | Counted output for a stated window with start and end timestamps, against the declared net ideal cycle time | A cycle time taken from the machine datasheet rather than from the observed distribution | | "Availability is Y%" | Every stop with start, end and cause code, captured at the machine during the run | Short stops that were never recorded, so availability is overstated and the gap shows up as an impossible performance figure | | "Maximum capacity is Z per week" | Shift pattern, planned downtime, changeover count and duration for the same period, plus the demonstrated rate | Maximum capacity computed at 100% availability, which no line achieves | | "We have capacity for your volume on a shared line" | Actual utilisation per shared asset by part number over the period, and the loading plan | Three customers each told they have the same press hours | ### How long the demonstration has to be Durations are customer-specific and the customer's own document is the only answer that counts. The published range is consistent in direction: hours to shifts, not minutes. Ficosa's Supplier Quality Manual (edition 6, 20 February 2026) requires the trial at the final manufacturing location with final resources, a minimum of two hours, demonstrating parts per hour, machine downtime, scrap levels, parts flow, material feed and in-process checks at actual production speed. Practitioner accounts describe a timed run at target cycle time with production tooling and staffing, typically a full shift, with demonstrated rate as net good parts divided by run duration over target cycle time; downtime, scrap and rework reduce the numerator and are not added back. Anything that makes the run unrepresentative invalidates it: an engineer standing at the machine, hand-picked material, extra staff. The German two-day trial often written as 2TP is asserted by consultancy sources, not by any OEM document we opened, so do not plan two days because a blog said two days. Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### Why a number reconstructed from a spreadsheet fails Not because spreadsheets are bad. Because a reconstructed capacity figure has a signature, and reviewers know it. - The arithmetic contradicts itself. The CAR manual treats a demonstrated OEE above 100% as an automatic red flag. That is what you get when the declared cycle time and the counted output do not match. - Stops are missing. Micro-stops nobody writes down inflate availability, performance absorbs the difference, and neither factor tells you anything any more. - The time window is vague. A capacity figure with no stated start and end cannot be tied to a shift pattern, a product mix or an equipment state, so it cannot be reproduced. - Nobody can answer the second question. Certification bodies report the same pattern across quality systems: the record exists, but there is no concrete case behind it. A capacity figure is a claim, and the follow-up is always what it was measured on. - The manual run-at-rate drifts too. Hand collection of demonstrated rate, downtime, scrap and good-part counts during the run is still common, and it creates accuracy gaps at the moment the number matters most. ### What we could not verify, and are not going to assert - Volkswagen Formel Q capacity analysis. The openly available Formel Q Capability, 9th edition, December 2022, contains no capacity, run-at-rate or capability clause; it routes process audits to VDA 6.3 elements P5 to P7, product audits to VDA 6.5, and escalation to the Critical Suppliers programme. The capacity method sits in Formel Q New Parts Integral, which is not published openly. Not verified. - BMW logistics and capacity requirements. The BMW CSR is deliberately short and carries no capacity, run-at-rate, process-capability or escalation clause. The operational logistics requirements sit behind the b2b.bmw.com login. Not verified. - Scania and CNH Industrial. Neither publishes a CSR on the IATF library; Scania's standards are login-only. Not verified. - SupplyOn as an OEM capacity channel. Its capacity assessment page names no OEM customers and lists no required data fields. Any claim that a named OEM collects capacity through SupplyOn needs separate verification. - Stellantis CAT mechanics, as above: a training provider's undated course description. ### Building a capacity file that survives the question - Fix the boundaryOne line or cell, named, with the part numbers it makes and the equipment it shares. Shared assets get a loading plan from the start, because IVECO clause 17 asks for shared lines by name. - Record the run, do not reconstruct itCounts, stops with cause codes and changeovers timed as they happen. The demonstration window and the evidence window have to be the same window. - Keep a rolling history, not a single trialThe CAR manual's 25 weeks is a reasonable target for any customer. A rolling weekly rate and OEE series answers the capacity question on the day it is asked, instead of triggering a special run. - Reconcile to what shippedDeclared capacity against produced quantity against delivered quantity, per period, with a recorded reason for any divergence. This is what makes a portal figure plausible rather than merely present. - Re-prove on changeProcess, equipment or location change resets the demonstration. Date the new run after the change and retire the old sheet rather than editing it. You can do the first three steps with a clipboard and a spreadsheet, and many plants should start there: it costs nothing and it tells you whether your cycle times are real. What breaks is the fourth step, once several lines and several customers are involved and the reconciliation has to happen every week. At that point plants move to automatic counting, whether through the MES they already own, an integrator's data layer, or a dedicated monitoring system. - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - IATF Rules 6th edition SI 5: a bad scorecard is now audit time - Machine capability studies: the conditions, not the maths - VDA 6.3 and production data: where the points go - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Does any OEM require a specific capacity software or portal?** Mercedes-Benz names its SMB portal modules BKM and BBM for capacity and for shortfall communication. Stellantis names GST, Amadeus/SQP and its other systems for performance data. Neither names a system you must run inside your own plant to produce the figures. IVECO asks for evidence of maintained capacity without specifying the tool. The requirement is a defensible number, not a particular vendor. **We run shared lines. What exactly has to be shown?** IVECO clause 17 covers dedicated and non-dedicated lines, so the capacity evidence has to exist for shared equipment too. In practice that means actual utilisation per shared asset broken down by part number over a stated period, plus the loading plan showing how the remaining hours are allocated. The Plastic Omnium CAR manual treats total allocation above 90% as a warning sign, which is a useful internal threshold whatever your customer asks. **How long does a run-at-rate have to be?** It depends on the customer, and the customer's own document is the only answer that counts. The published range runs from a minimum of two hours at the final location with final resources, in Ficosa's supplier quality manual, to a full shift in practitioner accounts. The two-day German trial often quoted as 2TP comes from consultancy sources, not from an OEM document we could open. **Our capacity figure came out above 100% OEE. Is that possible?** No. The Plastic Omnium CAR manual flags it as an automatic red. It means the declared cycle time and the counted output contradict each other: either the ideal cycle time is slower than the line actually runs, or parts are being counted that the time window cannot contain. Fix the cycle time against the observed distribution before submitting anything. **Does a capacity claim have to be re-proved during serial production?** Mercedes requires current delivery capacity to be kept up to date in the portal, so yes in substance. IVECO asks for evidence of maintaining required capacity, which is a present-tense obligation. For Stellantis, secondary sources describe the Capacity Assessment Tool being re-run during serial life whenever process, equipment or location changes; confirm the current rule with your Stellantis contact, because the CSR itself does not carry the method. ### Sources - IATF Global Oversight, Customer-Specific Requirements library (versions as listed 22 Sep 2026) - Mercedes-Benz Special Terms 2020, document ALD00001238, edition 2019-11 (DocMaster public portal) - Mercedes-Benz Special Terms 2025, document ALD00002454, edition 2025-04 (listed, not read) - Stellantis, CSR for use with IATF 16949, v1, June 2025 - Plastic Omnium Lighting, Capacity Analysis Report (CAR) Manual, LSM-SCM-403 rev. 05, 8 November 2022 - Ficosa, Supplier Quality Manual, I-CP-XX/XX-16 edition 6, 20 February 2026 - Odette International, Global MMOG/LE Introduction and Instructions v6.0, March 2023 Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # VDA 6.3 and production data: where P6 points get lost URL: https://industrialprocessai.com/we/vda-6-3-production-data/ Summary: VDA 6.3 4th edition, January 2023: what P6 asks about equipment capability, targets and quantities, and the data trail that keeps the points. Translations: de https://industrialprocessai.com/de/we/vda-6-3-produktionsdaten/, es https://industrialprocessai.com/es/we/vda-6-3-datos-de-produccion/, it https://industrialprocessai.com/it/we/vda-6-3-dati-produzione/, tr https://industrialprocessai.com/tr/we/vda-6-3-uretim-verisi/ HomeCompliance · VDA 6.3 process audit · status 22 September 2026 ## VDA 6.3 and production data: what P6 examines, and where the points go For quality managers, process engineers and plant directors preparing a VDA 6.3 process audit. What the 4th edition covers in P6, the recurring points-loss around untraceable production data, and what an auditable data trail actually looks like. In shortVDA 6.3 in its 4th edition of January 2023 has 59 questions, 17 of special significance, plus 35 for the potential analysis. The production-data pressure sits in P6: equipment capability and maintenance in P6.4, targets, data collection and deviation analysis in P6.5, planned against produced quantities in P6.6. The recurring loss is production data that cannot be traced to the characteristic. The TEEPTRAK approach Audit evidence that comes from the machines, not from spreadsheets. OEE, stop history and planned against produced quantities are easier to defend when they are recorded automatically and time-stamped. TEEPTRAK keeps that record for every machine and exports it for audits and customer requests. - OEE and stop history per machine, with timestamps - Recorded stops as the basis for MTBF and MTTR, instead of reconstructed logs - Exports ready for auditors and customer scorecards Used in 700+ factories. At Hutchinson, OEE rose from 47% to 72%. Installed with a module wired into the electrical cabinet and a sensor at the edge of the line. No PLC changes. TEEPTRAK is not an MES: it measures, so you know what to fix first. See how TEEPTRAK works →Get the evidence pack → What this is based on and what it is notVDA 6.3 is copyrighted and paywalled. No question, criterion or catalogue text is reproduced here, and none should be. What follows is a paraphrase of the audit's structure and of what practitioners and certification bodies report about it, written from sources that were opened and read. The standard prevails over this page in every case. Buy the current volume from VDA QMC before you build anything on it. ### Which edition applies, and why last year's score is not comparable The current edition is the 4th, published January 2023, superseding the 2016 third edition. As of 22 September 2026 the VDA QMC publications listing still shows the 4th edition of January 2023, so there is no newer version to move to. - 59 questions in the questionnaire, of which 17 are questions of special significance, commonly written as star questions. A further 35 questions serve the potential analysis. The star questions were partly reallocated in 2023, but the count stayed the same. - What changed in 2023: software topics folded into the questionnaire; harmonisation with Automotive SPICE and the VDA Maturity Level Assurance method; purchasing requirements added into P3 and P4; guidance for remote audits incorporated; the chapters on the audit process and audit services removed. - Scores are not comparable across editions. DQS states plainly that results under the 2023 edition are not directly comparable to results under 2016. If you benchmark supplier scores year over year, that break has to be in your data. Questions are scored on a discrete scale of 10, 8, 6, 4 or 0. There are no intermediate values, which is why a single weak answer moves a sub-element noticeably. ### Where production data actually gets examined The audit runs P1 to P7: potential analysis, project management, planning of product and process development, realisation of product and process development, supplier management, process analysis and series production, and customer care. P6 is the core element for anyone running a line, and three of its six sub-elements are about production data. P6 sub-elements and the production data behind them | Sub-element | Subject | Data it rests on | | P6.1 | Process input: project handover, material availability, storage, identification and records | Material lot identity carried through to the part | | P6.2 | Process management: control-plan requirements in practice, repeat release after interruptions, special characteristics, handling of defective parts, material flow | Interruption and restart events, and the release decision attached to each one | | P6.3 | Personnel resources: operator capability, responsibilities, availability | Who was logged on, and against which qualification | | P6.4 | Material resources: equipment and machine capability, maintenance, measuring equipment, workstations | Per-asset availability, stoppage events with causes, completed maintenance, calibration status | | P6.5 | Effectiveness and efficiency: targets and key figures, data collection, analysis of deviations, regular audits | Target against actual per line and shift, each deviation with a reason and a linked action | | P6.6 | Process output: quantities and batch sizes, storage and transport, records, customer requirements at delivery | Planned against produced quantity per order and batch, split into good, scrap and rework | Sub-element scope as described in openly published training material, August 2024. Paraphrase only; the standard prevails. P6.4 and P6.5 are the adjacent territory to IATF 16949 clause 8.5.1.5 on total productive maintenance, and the same records serve both. If you have built maintenance objectives for the IATF clause, see our page on IATF 16949 §8.5.1.5 and reuse them here rather than building a second set. P6.6 is the one people underestimate: it is about output rate and batch reality, not only conformity. ### The recurring points-loss: data that cannot be traced to the characteristic This is the finding that appears again and again in practitioner accounts of VDA 6.3 audits, and it is worth spelling out because the mechanism is mundane. The worked exampleThe control plan calls a bore diameter a critical characteristic. The measurement system stores the same value under a different name, with different limits. Nothing links the two. Proving that the characteristic was checked correctly turns into manual data matching in front of the auditor, and what the auditor sees is a system that cannot answer its own question. Auditors review SPC charts, capability values and inspection records, and expect a reaction plan when capability falls below requirement. All of that depends on being able to resolve a control-plan line to its readings. Where that resolution requires a person with local knowledge and half an hour, the evidence is weak even when the underlying work was done properly. One widely circulated practitioner article reports that roughly 75% of critical findings fall in P5 and P6, attributed to weak detection of and reaction to process deviations in real time and to records that exist on paper but cannot be backed by a concrete, verifiable case when the auditor asks. That source carries no author and no date, so treat the percentage as an impression rather than a statistic. The underlying observation matches what certification bodies report about IATF audits generally: the frequent findings are not "we did not do it", they are "we cannot show that we did it". Free tool OEM Capacity & Audit Evidence Pack Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.Get the evidence pack → ### What an auditable data trail looks like The test is whether a single question can be answered from the record without human reconstruction. Pick a part made three months ago and follow it. Auditor's question to the record that answers it without reconstruction | Question | What must exist | Sub-element | | Which machine made this part, when, and who was on it? | Part or batch identity resolving to machine, operation, time window and operator logon | P6.1, P6.3 | | Show me this characteristic being checked at the planned frequency last quarter | Measurement records keyed to the control-plan characteristic ID, with achieved frequency against planned | P6.2 | | What happened after the line was interrupted on that shift? | Interruption event, the repeat-release decision, and the disposition of parts made since the previous good check | P6.2 | | Was this asset maintained as planned? | Preventive maintenance planned against done, on schedule, per asset, computed rather than transcribed | P6.4 | | Show me a target that was missed and what you did | Target against actual per line and shift, each deviation carrying a recorded reason and a linked action with an owner and a date | P6.5 | | Planned quantity was X. What did you actually make? | Planned against produced per order and batch, split good, scrap and rework, reconcilable to the delivery note | P6.6 | Three properties make the difference between a record that survives this and one that does not. Timestamps written at the event, not entered at the end of the shift. Identifiers shared between systems, so the control plan, the measurement record and the production record use the same name for the same characteristic. A retention window that covers the audit period and is queryable during the audit rather than restorable afterwards. ### A/B/C thresholds: sources disagree, so check the standard The overall result is classified A, B or C. We are not publishing a threshold, because the secondary sources we could open do not agree and the standard itself is paywalled. - Two practitioner sources give A at 90% and above, B from 80 to 89%, C below 80%. - Another gives A from 90 to 100%, B from 75 to 89%, C below 75%. - The 80% boundary is the more commonly cited of the two. That is not the same as being correct. Take the threshold from your licensed copy of VDA 6.3 or from your customer's own requirement, and do not let a supplier development plan hang on a number from a blog. Downgrading rules also exist and can pull a result below what the arithmetic average alone would give: the 2016 rules were carried into the 2023 edition. Published descriptions of the exact mechanics vary and come from vendor commentary, so we are not restating them either. What is safe to plan around is the direction: a zero, and in some accounts a low score on a star question, hurts disproportionately, so a single unanswerable question in P6 is not averaged away. ### Which customers run VDA 6.3, and which run something else - Volkswagen conducts process audits per VDA 6.3 using elements P5 to P7, and product audits per VDA 6.5 at least annually, according to the openly published Formel Q Capability, 9th edition, December 2022. Escalation runs through the Critical Suppliers programme. - Renault does not carry a VDA 6.3 clause in its customer-specific requirements. It requires a minimum ASES rank instead, which makes ASES the operative process assessment. Published descriptions of the ASES structure and grading come from a 2018 practitioner blog and the scheme may have changed, so confirm the current method with Renault rather than with us. - Tier 1 cascades matter. Ficosa escalates a supplier to its Level 1 on four or more non-A months out of six, two consecutive C ratings, or a C rating in a VDA 6.3 audit. A process-audit grade can therefore trigger a commercial escalation directly. - What an OEE system really costs: the 3-year calculator - IATF 16949 §8.5.1.5 TPM: OEE, MTBF and MTTR audit evidence - Machine capability studies: the conditions, not the maths - IATF Rules 6th edition SI 5: a bad scorecard is now audit time - OEM capacity evidence: the standing, portal-reported obligation - OEM Capacity and Audit Evidence Pack (xlsx) ### Questions **Which VDA 6.3 edition applies in 2026?** The 4th edition, published January 2023. The VDA QMC publications listing as of 22 September 2026 still shows it as current. Results under the 2023 edition are not directly comparable to results under the 2016 edition, so any year-over-year supplier benchmark that crosses the change needs a note on it. **How many questions and star questions are there?** 59 questions in the questionnaire, of which 17 are questions of special significance, plus 35 questions for the potential analysis. The star questions were partly reallocated in the 2023 revision but the count did not change. The questions themselves are copyrighted and are not reproduced here or anywhere else legitimate. **What is the A/B/C threshold?** We are not publishing one. The sources we could open disagree: some give B as 80 to 89% and C below 80%, another gives B as 75 to 89% and C below 75%. The 80% boundary is more commonly cited but that does not settle it. Take the threshold from your licensed copy of the standard or from the customer requirement that is driving the audit. **Does VDA 6.3 require an MES or real-time data capture?** No standard we examined mandates a particular system. What P6 asks for is evidence: equipment capability and maintenance records, targets with collected data and deviation analysis, and planned against produced quantities. A plant can produce all of that on paper if the paper is complete, timely and traceable. The reason plants move to automatic capture is that the traceability and the timeliness get hard to sustain by hand, not that a standard demands it. **Our records are complete but stored in three systems. Is that a problem?** It becomes one when the systems use different identifiers for the same characteristic or operation. The recurring points-loss reported by practitioners is production data that cannot be traced back to the relevant characteristic, and the usual mechanism is exactly this: a control plan and a measurement system that name the same thing differently, with no link between them. Fixing the identifiers is usually cheaper than replacing the systems. ### Sources - VDA QMC publications listing (VDA 6.3 Process Audit, 4th edition, January 2023), accessed 22 September 2026 - DQS, "Changes to VDA 6.3:2023", 22 February 2023 - Volkswagen AG Group Supply Chain, Formel Q: Capability, 9th edition, December 2022 (openly hosted copy) - IATF Global Oversight, Customer-Specific Requirements library (Renault v5.0, April 2026) - Ficosa, Supplier Quality Manual, I-CP-XX/XX-16 edition 6, 20 February 2026 - Smithers, "Top 5 IATF 16949 Non-Conformances via the IAOB", December 2025 Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit. --- # Translated pages by language ## Français - [Boîte à outils de l'acheteur d'IA industrielle: grille, questions RFP, diagnostics](https://industrialprocessai.com/fr/) - [Coût d’un logiciel TRS : le calculateur du coût total sur 3 ans](https://industrialprocessai.com/fr/cout-logiciel-trs/) - [Grille d'évaluation des fournisseurs d'IA industrielle + RFP: téléchargement gratuit](https://industrialprocessai.com/fr/get/) - [Guides IA industrielle](https://industrialprocessai.com/fr/guides/) - [12 exemples réels d'IA dans l'industrie qui produisent du ROI (2026)](https://industrialprocessai.com/fr/guides/ai-manufacturing-examples/) - [IA générative dans l'industrie: ce qui marche, ce qui ne marche pas (2026)](https://industrialprocessai.com/fr/guides/generative-ai-manufacturing/) - [Qu'est-ce qu'une plateforme d'IA industrielle ? Définition, catégories, ROI (2026)](https://industrialprocessai.com/fr/guides/what-is-industrial-ai/) - [Les six critères pour évaluer une plateforme d'IA industrielle](https://industrialprocessai.com/fr/methodology/) - [Grille d'achat : plateformes d'IA industrielle (Excel gratuit)](https://industrialprocessai.com/fr/scorecard/) - [Capabilité machine: les conditions qui rendent le Cmk valable](https://industrialprocessai.com/fr/we/capabilite-machine-conditions/) - [Énergie par pièce: ce que les clients demandent en 2026](https://industrialprocessai.com/fr/we/energie-par-piece/) - [IATF Rules SI 5 : un mauvais scorecard OEM coûte de l'audit](https://industrialprocessai.com/fr/we/iatf-rules-si5/) - [Pack preuves capacité OEM et audit (xlsx): gratuit](https://industrialprocessai.com/fr/we/pack-preuves/) - [Preuves de capacité OEM : ce qu'un fournisseur doit prouver](https://industrialprocessai.com/fr/we/preuves-capacite-oem/) - [Preuves ISO 50001: IPE, référence et plan de mesure](https://industrialprocessai.com/fr/we/preuves-iso-50001/) ## Deutsch - [Einkaufs-Toolkit für Industrie-KI: Scorecard, RFP-Fragen, Readiness-Checks](https://industrialprocessai.com/de/) - [Bewertungsraster für Industrie-KI-Anbieter + RFP: kostenloser Download](https://industrialprocessai.com/de/get/) - [Industrie-KI-Leitfäden](https://industrialprocessai.com/de/guides/) - [12 reale Beispiele für KI in der Fertigung mit echtem ROI (2026)](https://industrialprocessai.com/de/guides/ai-manufacturing-examples/) - [Generative KI in der Fertigung: Was funktioniert, was nicht (2026)](https://industrialprocessai.com/de/guides/generative-ai-manufacturing/) - [Was ist eine industrielle KI-Plattform? Definition, Kategorien, ROI (2026)](https://industrialprocessai.com/de/guides/what-is-industrial-ai/) - [Die sechs Kriterien zur Bewertung einer Industrie-KI-Plattform](https://industrialprocessai.com/de/methodology/) - [OEE-Software Kosten: der Gesamtkostenrechner für 3 Jahre (2026)](https://industrialprocessai.com/de/oee-software-kosten/) - [Einkaufs-Scorecard: industrielle KI-Plattformen (kostenloses Excel)](https://industrialprocessai.com/de/scorecard/) - [Energie pro Teil: was Kunden 2026 von Zulieferern fordern](https://industrialprocessai.com/de/we/energie-pro-teil/) - [IATF Rules SI 5: schlechte OEM-Scorecard kostet Auditzeit](https://industrialprocessai.com/de/we/iatf-rules-si5/) - [ISO 50001 Nachweise: EnPIs, Baselines und der Messplan](https://industrialprocessai.com/de/we/iso-50001-nachweise/) - [Maschinenfähigkeit: die Bedingungen, die Cmk tragen](https://industrialprocessai.com/de/we/maschinenfaehigkeit-bedingungen/) - [Nachweispaket OEM-Kapazität und Audit (xlsx): gratis](https://industrialprocessai.com/de/we/nachweispaket/) - [OEM-Kapazitätsnachweis: was Zulieferer 2026 belegen müssen](https://industrialprocessai.com/de/we/oem-kapazitaetsnachweis/) - [VDA 6.3 und Produktionsdaten: wo P6-Punkte verloren gehen](https://industrialprocessai.com/de/we/vda-6-3-produktionsdaten/) ## Español - [Kit del comprador de IA industrial: scorecard, preguntas RFP, diagnósticos](https://industrialprocessai.com/es/) - [Coste de un software OEE: calculadora del coste total a 3 años](https://industrialprocessai.com/es/coste-software-oee/) - [Scorecard de evaluación de proveedores de IA industrial + RFP: descarga gratuita](https://industrialprocessai.com/es/get/) - [Guías de IA industrial](https://industrialprocessai.com/es/guides/) - [12 ejemplos reales de IA en la industria que producen ROI (2026)](https://industrialprocessai.com/es/guides/ai-manufacturing-examples/) - [IA generativa en la industria: lo que funciona y lo que no (2026)](https://industrialprocessai.com/es/guides/generative-ai-manufacturing/) - [¿Qué es una plataforma de IA industrial? Definición, categorías, ROI (2026)](https://industrialprocessai.com/es/guides/what-is-industrial-ai/) - [Los seis criterios para evaluar una plataforma de IA industrial](https://industrialprocessai.com/es/methodology/) - [Scorecard de compra: plataformas de IA industrial (Excel gratis)](https://industrialprocessai.com/es/scorecard/) - [Capacidad de máquina: las condiciones que sostienen el Cmk](https://industrialprocessai.com/es/we/capacidad-de-maquina-condiciones/) - [Energía por pieza: qué piden los clientes a sus proveedores](https://industrialprocessai.com/es/we/energia-por-pieza/) - [Evidencia de capacidad OEM: qué debe probar un proveedor](https://industrialprocessai.com/es/we/evidencia-de-capacidad-oem/) - [Evidencias ISO 50001: IDEn, líneas base y plan de medición](https://industrialprocessai.com/es/we/iso-50001-evidencias/) - [Pack de evidencias de capacidad y auditoría OEM (xlsx)](https://industrialprocessai.com/es/we/pack-de-evidencias/) - [IATF Rules SI 5: un mal scorecard OEM cuesta horas de auditoría](https://industrialprocessai.com/es/we/reglas-iatf-si-5/) - [VDA 6.3 y datos de producción: dónde se pierden los puntos P6](https://industrialprocessai.com/es/we/vda-6-3-datos-de-produccion/) ## Italiano - [Kit dell'acquirente di AI industriale: scorecard, domande RFP, verifiche di prontezza](https://industrialprocessai.com/it/) - [Costo di un software OEE: il calcolatore del costo totale a 3 anni](https://industrialprocessai.com/it/costo-software-oee/) - [Valutazione della prontezza dei dati per l'AI industriale](https://industrialprocessai.com/it/data-readiness/) - [Scorecard di valutazione dei fornitori di AI industriale + RFP: Download gratuito](https://industrialprocessai.com/it/get/) - [Guide all'AI industriale](https://industrialprocessai.com/it/guides/) - [12 esempi reali di AI nella produzione che producono ROI (2026)](https://industrialprocessai.com/it/guides/ai-manufacturing-examples/) - [AI generativa nella produzione: cosa funziona e cosa no (2026)](https://industrialprocessai.com/it/guides/generative-ai-manufacturing/) - [Cos'è una piattaforma di AI industriale? Definizione, categorie, ROI (guida 2026)](https://industrialprocessai.com/it/guides/what-is-industrial-ai/) - [I sei criteri per valutare una piattaforma di AI industriale](https://industrialprocessai.com/it/methodology/) - [Revisione di passaggio dal pilota alla produzione](https://industrialprocessai.com/it/pilot-gate-review/) - [Banca di domande RFP per l'AI industriale](https://industrialprocessai.com/it/rfp-questions/) - [Scorecard dell'acquirente di piattaforme di AI industriale (xlsx gratuito)](https://industrialprocessai.com/it/scorecard/) - [Capability macchina: le condizioni che rendono valido il Cmk](https://industrialprocessai.com/it/we/capability-macchina-condizioni/) - [Energia per pezzo: cosa chiedono i clienti ai fornitori nel 2026](https://industrialprocessai.com/it/we/energia-per-pezzo/) - [Evidenze di capacità OEM: cosa deve dimostrare un fornitore](https://industrialprocessai.com/it/we/evidenze-capacita-oem/) - [Evidenze ISO 50001: EnPI, baseline e piano di misura](https://industrialprocessai.com/it/we/evidenze-iso-50001/) - [Pack evidenze capacità OEM e audit (xlsx): download gratuito](https://industrialprocessai.com/it/we/pack-evidenze/) - [IATF Rules SI 5: una scorecard negativa costa ore di audit](https://industrialprocessai.com/it/we/regole-iatf-si5/) - [VDA 6.3 e dati di produzione: dove si perdono i punti in P6](https://industrialprocessai.com/it/we/vda-6-3-dati-produzione/) ## Nederlands - [Industriële-AI-toolkit voor kopers: scorecard, RFP-vragen, readiness-checks](https://industrialprocessai.com/nl/) - [Scorecard voor industriële AI-leveranciers + RFP: gratis download](https://industrialprocessai.com/nl/get/) - [Industriële AI-gidsen](https://industrialprocessai.com/nl/guides/) - [12 echte voorbeelden van AI in de industrie die ROI opleveren (2026)](https://industrialprocessai.com/nl/guides/ai-manufacturing-examples/) - [Generatieve AI in de industrie: wat werkt en wat niet (2026)](https://industrialprocessai.com/nl/guides/generative-ai-manufacturing/) - [Wat is een industrieel AI-platform? Definitie, categorieën, ROI (2026)](https://industrialprocessai.com/nl/guides/what-is-industrial-ai/) - [Kosten van OEE-software: rekentool voor de totale kosten over 3 jaar](https://industrialprocessai.com/nl/kosten-oee-software/) - [De zes criteria om een industrieel AI-platform te beoordelen](https://industrialprocessai.com/nl/methodology/) - [Inkoop-scorecard: industriële AI-platformen (gratis Excel)](https://industrialprocessai.com/nl/scorecard/) - [Energie per stuk: wat klanten leveranciers vragen in 2026](https://industrialprocessai.com/nl/we/energie-per-stuk/) - [ISO 50001-bewijs: EnPI's, baselines en het meetplan](https://industrialprocessai.com/nl/we/iso-50001-bewijs/) ## Português - [Kit do comprador de IA industrial: grelha de avaliação, perguntas RFP, verificações de prontidão](https://industrialprocessai.com/pt/) - [Custo de um software OEE: calculadora do custo total a 3 anos](https://industrialprocessai.com/pt/custo-software-oee/) - [Avaliação da prontidão dos dados para IA industrial](https://industrialprocessai.com/pt/data-readiness/) - [Grelha de avaliação de fornecedores de IA industrial + RFP: descarga gratuita](https://industrialprocessai.com/pt/get/) - [Guias de IA industrial](https://industrialprocessai.com/pt/guides/) - [12 exemplos reais de IA na indústria que produzem ROI (2026)](https://industrialprocessai.com/pt/guides/ai-manufacturing-examples/) - [IA generativa na indústria: o que funciona e o que não funciona (2026)](https://industrialprocessai.com/pt/guides/generative-ai-manufacturing/) - [O que é uma plataforma de IA industrial? Definição, categorias, ROI (guia 2026)](https://industrialprocessai.com/pt/guides/what-is-industrial-ai/) - [Os seis critérios para avaliar uma plataforma de IA industrial](https://industrialprocessai.com/pt/methodology/) - [Revisão de passagem do piloto à produção](https://industrialprocessai.com/pt/pilot-gate-review/) - [Banco de perguntas RFP para IA industrial](https://industrialprocessai.com/pt/rfp-questions/) - [Grelha do comprador de plataformas de IA industrial (xlsx gratuito)](https://industrialprocessai.com/pt/scorecard/) ## Polski - [Zestaw kupującego AI w przemyśle: karta oceny, pytania do RFP, testy gotowości](https://industrialprocessai.com/pl/) - [IATF 16949 §8.5.1.5 TPM: dowody audytowe dla OEE, MTBF i MTTR](https://industrialprocessai.com/pl/cee/iatf-16949-tpm-oee/) - [AI w przemyśle CEE: dlaczego warstwa danych musi być pierwsza](https://industrialprocessai.com/pl/cee/najpierw-dane-potem-ai/) - [Pakiet dowodów IATF TPM: arkusz OEE, MTBF, MTTR (xlsx)](https://industrialprocessai.com/pl/cee/pakiet-dowodow-iatf-tpm/) - [Zapytania OEM do dostawców: energia na sztukę i OEE po Omnibusie](https://industrialprocessai.com/pl/cee/zapytania-oem-dane-dostawcow/) - [Ocena gotowości danych do AI w przemyśle](https://industrialprocessai.com/pl/data-readiness/) - [Karta oceny dostawców AI dla przemysłu + RFP: bezpłatne pobranie](https://industrialprocessai.com/pl/get/) - [Przewodniki po AI w przemyśle](https://industrialprocessai.com/pl/guides/) - [12 prawdziwych przykładów AI w produkcji, które dają ROI (2026)](https://industrialprocessai.com/pl/guides/ai-manufacturing-examples/) - [Generatywna AI w produkcji: co działa, a co nie (2026)](https://industrialprocessai.com/pl/guides/generative-ai-manufacturing/) - [Czym jest platforma AI dla przemysłu? Definicja, kategorie, ROI (przewodnik 2026)](https://industrialprocessai.com/pl/guides/what-is-industrial-ai/) - [Ile kosztuje system OEE? Kalkulator kosztów monitoringu produkcji](https://industrialprocessai.com/pl/ile-kosztuje-system-oee/) - [Sześć kryteriów oceny platformy AI dla przemysłu](https://industrialprocessai.com/pl/methodology/) - [Przegląd bramki od pilotażu do produkcji](https://industrialprocessai.com/pl/pilot-gate-review/) - [Baza pytań do RFP dla AI w przemyśle](https://industrialprocessai.com/pl/rfp-questions/) - [Karta oceny kupującego platformy AI dla przemysłu (darmowy xlsx)](https://industrialprocessai.com/pl/scorecard/) - [Zdolność produkcyjna: czego OEM wymaga od dostawcy w 2026](https://industrialprocessai.com/pl/we/zdolnosc-produkcyjna-oem/) ## Čeština - [Sada kupujícího průmyslovou AI: hodnoticí karta, otázky do RFP, kontroly připravenosti](https://industrialprocessai.com/cs/) - [Balíček důkazů TPM pro IATF: sledování OEE, MTBF, MTTR (xlsx)](https://industrialprocessai.com/cs/cee/balicek-dukazu-tpm-iatf/) - [AI ve výrobě ve střední a východní Evropě: nejdřív datová vrstva](https://industrialprocessai.com/cs/cee/data-pred-ai/) - [IATF 16949 §8.5.1.5 TPM: důkazy k OEE, MTBF a MTTR pro audit](https://industrialprocessai.com/cs/cee/iatf-16949-tpm-oee/) - [Požadavky OEM na data dodavatelů: energie na díl a OEE](https://industrialprocessai.com/cs/cee/pozadavky-oem-na-data-dodavatelu/) - [Kolik stojí monitoring výroby a měření OEE: kalkulačka na 3 roky](https://industrialprocessai.com/cs/cena-monitoringu-vyroby/) - [Hodnocení připravenosti dat pro průmyslovou AI](https://industrialprocessai.com/cs/data-readiness/) - [Hodnoticí karta dodavatelů průmyslové AI + RFP: ke stažení zdarma](https://industrialprocessai.com/cs/get/) - [Návody k průmyslové AI](https://industrialprocessai.com/cs/guides/) - [12 skutečných příkladů AI ve výrobě, které přinášejí ROI (2026)](https://industrialprocessai.com/cs/guides/ai-manufacturing-examples/) - [Generativní AI ve výrobě: co funguje a co ne (2026)](https://industrialprocessai.com/cs/guides/generative-ai-manufacturing/) - [Co je platforma průmyslové AI? Definice, kategorie, ROI (návod 2026)](https://industrialprocessai.com/cs/guides/what-is-industrial-ai/) - [Šest kritérií pro hodnocení platformy průmyslové AI](https://industrialprocessai.com/cs/methodology/) - [Revize brány od pilotu do výroby](https://industrialprocessai.com/cs/pilot-gate-review/) - [Databáze otázek do RFP pro průmyslovou AI](https://industrialprocessai.com/cs/rfp-questions/) - [Hodnoticí karta kupujícího platformy průmyslové AI (xlsx zdarma)](https://industrialprocessai.com/cs/scorecard/) ## Slovenčina - [Súprava kupujúceho priemyselnú AI: hodnotiaca karta, otázky do RFP, kontroly pripravenosti](https://industrialprocessai.com/sk/) - [Balík dôkazov IATF TPM: sledovanie OEE, MTBF, MTTR (xlsx)](https://industrialprocessai.com/sk/cee/balik-dokazov-iatf-tpm/) - [AI vo výrobe v SVE: prečo musí byť dátová vrstva na prvom mieste](https://industrialprocessai.com/sk/cee/data-pred-ai/) - [Dátové požiadavky OEM: energia na diel a OEE po Omnibuse](https://industrialprocessai.com/sk/cee/datove-poziadavky-oem/) - [IATF 16949 §8.5.1.5 TPM: dôkazy OEE, MTBF a MTTR pre audit](https://industrialprocessai.com/sk/cee/iatf-16949-tpm-oee/) - [Koľko stojí monitoring výroby a meranie OEE: kalkulačka na 3 roky](https://industrialprocessai.com/sk/cena-monitoringu-vyroby/) - [Hodnotenie pripravenosti dát pre priemyselnú AI](https://industrialprocessai.com/sk/data-readiness/) - [Hodnotiaca karta dodávateľov priemyselnej AI + RFP: na stiahnutie zadarmo](https://industrialprocessai.com/sk/get/) - [Návody k priemyselnej AI](https://industrialprocessai.com/sk/guides/) - [12 skutočných príkladov AI vo výrobe, ktoré prinášajú ROI (2026)](https://industrialprocessai.com/sk/guides/ai-manufacturing-examples/) - [Generatívna AI vo výrobe: čo funguje a čo nie (2026)](https://industrialprocessai.com/sk/guides/generative-ai-manufacturing/) - [Čo je platforma priemyselnej AI? Definícia, kategórie, ROI (návod 2026)](https://industrialprocessai.com/sk/guides/what-is-industrial-ai/) - [Šesť kritérií na hodnotenie platformy priemyselnej AI](https://industrialprocessai.com/sk/methodology/) - [Revízia brány od pilotu do výroby](https://industrialprocessai.com/sk/pilot-gate-review/) - [Databáza otázok do RFP pre priemyselnú AI](https://industrialprocessai.com/sk/rfp-questions/) - [Hodnotiaca karta kupujúceho platformy priemyselnej AI (xlsx zadarmo)](https://industrialprocessai.com/sk/scorecard/) ## Magyar - [Ipari MI-vásárlói eszköztár: értékelőlap, RFP-kérdések, felkészültségi ellenőrzések](https://industrialprocessai.com/hu/) - [MI a régió feldolgozóiparában: miért az adatréteg az első](https://industrialprocessai.com/hu/cee/adatok-az-mi-elott/) - [IATF 16949 §8.5.1.5 TPM: OEE, MTBF és MTTR auditbizonyítékok](https://industrialprocessai.com/hu/cee/iatf-16949-tpm-oee/) - [IATF TPM bizonyítékcsomag: OEE-, MTBF-, MTTR-követő (xlsx)](https://industrialprocessai.com/hu/cee/iatf-tpm-bizonyitek-csomag/) - [OEM beszállítói adatkérések: darabonkénti energia és OEE](https://industrialprocessai.com/hu/cee/oem-beszallitoi-adatkeresek/) - [Ipari MI adatfelkészültségi felmérés](https://industrialprocessai.com/hu/data-readiness/) - [Ipari MI-szállítói értékelőlap + RFP: ingyenes letöltés](https://industrialprocessai.com/hu/get/) - [Ipari MI-útmutatók](https://industrialprocessai.com/hu/guides/) - [12 valós példa a mesterséges intelligenciára a gyártásban, amely megtérül (2026)](https://industrialprocessai.com/hu/guides/ai-manufacturing-examples/) - [Generatív MI a gyártásban: mi működik és mi nem (2026)](https://industrialprocessai.com/hu/guides/generative-ai-manufacturing/) - [Mi az ipari MI-platform? Definíció, kategóriák, megtérülés (2026-os útmutató)](https://industrialprocessai.com/hu/guides/what-is-industrial-ai/) - [Az ipari MI-platform értékelésének hat szempontja](https://industrialprocessai.com/hu/methodology/) - [Mennyibe kerül az OEE mérés? Gyártásfelügyeleti rendszer 3 évre](https://industrialprocessai.com/hu/oee-meres-koltsege/) - [Pilotból éles termelésbe: kapuellenőrzés](https://industrialprocessai.com/hu/pilot-gate-review/) - [Ipari MI RFP-kérdésbank](https://industrialprocessai.com/hu/rfp-questions/) - [Ipari MI-platform vásárlói értékelőlap (ingyenes xlsx)](https://industrialprocessai.com/hu/scorecard/) ## Română - [Trusa cumpărătorului de AI industrial: fișă de evaluare, întrebări RFP, verificări](https://industrialprocessai.com/ro/) - [IA în industria din ECE: de ce stratul de date vine primul](https://industrialprocessai.com/ro/cee/datele-inaintea-ia/) - [IATF 16949 §8.5.1.5 TPM: dovezi de audit pentru OEE, MTBF, MTTR](https://industrialprocessai.com/ro/cee/iatf-16949-tpm-oee/) - [Pachet de dovezi TPM IATF: evidență OEE, MTBF, MTTR (xlsx)](https://industrialprocessai.com/ro/cee/pachet-dovezi-tpm-iatf/) - [Solicitări de date OEM: energie pe piesă și OEE după Omnibus](https://industrialprocessai.com/ro/cee/solicitari-date-oem-furnizori/) - [Cât costă un software OEE? Calculatorul costului total pe 3 ani](https://industrialprocessai.com/ro/cost-software-oee/) - [Scorecard de evaluare furnizori AI industrial + RFP: descărcare gratuită](https://industrialprocessai.com/ro/get/) - [Ghiduri de AI industrial](https://industrialprocessai.com/ro/guides/) - [12 exemple reale de AI în industrie care produc ROI (2026)](https://industrialprocessai.com/ro/guides/ai-manufacturing-examples/) - [AI generativ în industrie: ce funcționează, ce nu (2026)](https://industrialprocessai.com/ro/guides/generative-ai-manufacturing/) - [Ce este o platformă de AI industrial? Definiție, categorii, ROI (2026)](https://industrialprocessai.com/ro/guides/what-is-industrial-ai/) - [Cele șase criterii pentru evaluarea unei platforme de AI industrial](https://industrialprocessai.com/ro/methodology/) - [Fișa de evaluare: platforme de AI industrial (Excel gratuit)](https://industrialprocessai.com/ro/scorecard/) - [Dovezi de capacitate OEM: ce trebuie să arate furnizorii în 2026](https://industrialprocessai.com/ro/we/dovezi-capacitate-oem/) ## Svenska - [Verktygslåda för köpare av industriell AI: poängkort, RFP-frågor, beredskapskontroller](https://industrialprocessai.com/sv/) - [Bedömning av databeredskap för industriell AI](https://industrialprocessai.com/sv/data-readiness/) - [Poängkort för utvärdering av leverantörer av industriell AI + RFP: gratis nedladdning](https://industrialprocessai.com/sv/get/) - [Guider om industriell AI](https://industrialprocessai.com/sv/guides/) - [12 verkliga exempel på AI i tillverkning som ger avkastning (2026)](https://industrialprocessai.com/sv/guides/ai-manufacturing-examples/) - [Generativ AI i tillverkning: vad som fungerar och inte (2026)](https://industrialprocessai.com/sv/guides/generative-ai-manufacturing/) - [Vad är en plattform för industriell AI? Definition, kategorier, avkastning (guide 2026)](https://industrialprocessai.com/sv/guides/what-is-industrial-ai/) - [Vad kostar ett OEE-system? Kalkylator för totalkostnaden på 3 år](https://industrialprocessai.com/sv/kostnad-oee-system/) - [De sex kriterierna för att poängsätta en plattform för industriell AI](https://industrialprocessai.com/sv/methodology/) - [Grindgranskning från pilot till produktion](https://industrialprocessai.com/sv/pilot-gate-review/) - [Frågebank för RFP om industriell AI](https://industrialprocessai.com/sv/rfp-questions/) - [Köparens poängkort för plattformar för industriell AI (gratis xlsx)](https://industrialprocessai.com/sv/scorecard/) - [Energi per detalj: vad kunderna begär av leverantörer 2026](https://industrialprocessai.com/sv/we/energi-per-detalj/) - [ISO 50001-bevis: EnPI, baslinjer och mätplanen](https://industrialprocessai.com/sv/we/iso-50001-bevis/) ## Türkçe - [Endüstriyel yapay zekâ alıcı araç seti: değerlendirme kartı, RFP soruları, hazırlık kontrolleri](https://industrialprocessai.com/tr/) - [Endüstriyel yapay zekâ veri hazırlık değerlendirmesi](https://industrialprocessai.com/tr/data-readiness/) - [Endüstriyel yapay zekâ tedarikçi değerlendirme kartı + RFP: ücretsiz indirme](https://industrialprocessai.com/tr/get/) - [Endüstriyel yapay zekâ rehberleri](https://industrialprocessai.com/tr/guides/) - [Üretimde yatırım getirisi sağlayan 12 gerçek yapay zekâ örneği (2026)](https://industrialprocessai.com/tr/guides/ai-manufacturing-examples/) - [Üretimde üretken yapay zekâ: ne işe yarar, ne yaramaz (2026)](https://industrialprocessai.com/tr/guides/generative-ai-manufacturing/) - [Endüstriyel yapay zekâ platformu nedir? Tanım, kategoriler, yatırım getirisi (2026 rehberi)](https://industrialprocessai.com/tr/guides/what-is-industrial-ai/) - [Endüstriyel yapay zekâ platformunu puanlamak için altı ölçüt](https://industrialprocessai.com/tr/methodology/) - [Pilottan üretime geçiş incelemesi](https://industrialprocessai.com/tr/pilot-gate-review/) - [Endüstriyel yapay zekâ RFP soru bankası](https://industrialprocessai.com/tr/rfp-questions/) - [Endüstriyel yapay zekâ platformu alıcı değerlendirme kartı (ücretsiz xlsx)](https://industrialprocessai.com/tr/scorecard/) - [Üretim takip programı fiyatları: OEE sisteminin 3 yıllık maliyeti](https://industrialprocessai.com/tr/uretim-takip-programi-fiyatlari/) - [IATF Rules SI 5: kötü OEM karnesi artık denetim süresi demek](https://industrialprocessai.com/tr/we/iatf-rules-si5/) - [ISO 50001 kanıtı: EnPI'ler, temel çizgiler ve ölçüm planı](https://industrialprocessai.com/tr/we/iso-50001-kaniti/) - [OEM Kapasite ve Denetim Kanıt Paketi (xlsx): ücretsiz](https://industrialprocessai.com/tr/we/kanit-paketi/) - [Makine yeterliliği çalışması: Cmk'yı geçerli kılan koşullar](https://industrialprocessai.com/tr/we/makine-yeterliligi-kosullari/) - [OEM kapasite kanıtı: tedarikçiler 2026'da neyi kanıtlamalı](https://industrialprocessai.com/tr/we/oem-kapasite-kaniti/) - [Parça başına enerji: müşteriler 2026'da ne istiyor](https://industrialprocessai.com/tr/we/parca-basina-enerji/) - [VDA 6.3 ve üretim verisi: P6 puanları nerede kayboluyor](https://industrialprocessai.com/tr/we/vda-6-3-uretim-verisi/) ## 中文 - [工业AI采购工具包:评分表、RFP问题、就绪度检查](https://industrialprocessai.com/cn/) - [工业 AI 供应商评估评分卡 + RFP : 免费下载](https://industrialprocessai.com/cn/get/) - [工业AI指南](https://industrialprocessai.com/cn/guides/) - [12个真实产生ROI的制造业AI案例(2026)](https://industrialprocessai.com/cn/guides/ai-manufacturing-examples/) - [制造业中的生成式AI:什么有效,什么无效(2026)](https://industrialprocessai.com/cn/guides/generative-ai-manufacturing/) - [什么是工业AI平台?定义、分类、ROI(2026)](https://industrialprocessai.com/cn/guides/what-is-industrial-ai/) - [评估工业AI平台的六项指标](https://industrialprocessai.com/cn/methodology/) - [OEE系统的真实成本:三年总成本计算器(2026)](https://industrialprocessai.com/cn/oee-system-cost/) - [选型评分表:工业AI平台(免费Excel)](https://industrialprocessai.com/cn/scorecard/)