Western Europe & Türkiye · Updated 22 September 2026
IndustrialProcessAI
Compliance · 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.
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
ClauseWhat it doesWhat has to exist
6.3 Energy reviewSystematic analysis of energy use and consumption; identifies significant energy uses and improvement opportunitiesThe 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 indicatorsThe metrics used to monitor performance and demonstrate improvementThe indicator list, the method for determining and updating each one, and the periodic values
6.5 Energy baselineThe quantitative reference each indicator is compared againstThe 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 dataThe measurement planWhat is metered, where, at what interval, with what accuracy, and how the data is stored
7.5 Documented informationCreation, updating and controlMeter identity, calibration status, and the version history of the method documents
9.1 Monitoring, measurement, analysis and evaluationEvaluation of energy performance and of the EnMSIndicator 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.

  1. 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.
  2. Write the revision conditions in advance. Guidance calls for a predetermined revision methodology, not a decision made when the numbers become inconvenient.
  3. 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:

  1. 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.
  2. 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.
  3. 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.
  4. 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
LayerIntervalWhy
Machine or cellOne minute or finerAnything coarser cannot separate producing from idle, and cannot attribute energy to a cycle, a part or a batch
Feeder or area15 minutesMatches the European settlement interval, so it reconciles with the utility bill and supports load-profile analysis
Reporting layerMonthlyIndicator 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
TriggerDutyDeadline
Above 85 TJ average annual final energy consumptionEnergy management system certified by an independent body; ISO 50001 is the standard the Commission guidance points to11 October 2027
Above 10 TJ and no EnMSEnergy audit, repeated at least every four yearsFirst audit 11 October 2026
Newly obliged enterprises crossing a threshold laterTwo years to put the EnMS in place, one year to complete the auditFrom 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.

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

  1. ISO 50001:2018, Energy management systems — Requirements with guidance for use
  2. ISO 50006:2023, Evaluating energy performance using energy performance indicators and energy baselines
  3. ISO 50015:2014, Measurement and verification of energy performance of organizations
  4. US DOE / LBNL, 50001 Ready Navigator, Task 11: Energy Performance Indicators and Energy Baselines
  5. Commission Recommendation (EU) 2024/2002 of 24 July 2024 on Article 11 of Directive (EU) 2023/1791
  6. European Commission, Energy Efficiency Directive overview (transposition deadline 11 October 2025)
  7. Denmark, Bekendtgørelse nr. 1138 of 18 September 2025, in force 11 October 2025
  8. Germany, Bundestag Drucksache 21/8027, 16 September 2026 (EnEfG/EDL-G reform bill, first reading)
  9. BAFA, Energieaudit (EnEfG and EDL-G duties in force), read 22 September 2026
  10. Energiavirasto, new energy efficiency obligations under act 1372/2025, news item 3 June 2026

Published by TEEPTRAK SAS, which makes production-monitoring and OEE software. Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit.