Energy KPIs: kWh/Part, CO₂/Part & CSRD Reporting

What energy KPIs actually measure — and why a number that was optional in 2022 is a customer-retention issue in 2026

Energy KPIs — principally kWh per part and CO₂ per part, together with their per-good-part variants and their aggregations to product family, line, shift and plant — translate energy consumption and associated emissions into a figure that can be compared, audited and defended. The arithmetic is simple: kWh consumed over a period divided by parts produced in the same period, then multiplied by the applicable emission factor per energy carrier for the CO₂ variant. In 2022 these numbers were an engineering curiosity maintained by the energy manager in a monthly spreadsheet. In 2026 they are a condition of business for every mid-market manufacturer supplying an EU-listed OEM, an audit line-item under the ESRS E1 standard for companies inside the CSRD Wave 2 scope, and — for energy-intensive exports into the EU — an input to the CBAM border-adjustment calculation. The question "what does this part cost us in kWh and in grams of CO₂?" has moved from the sustainability department's annual report into the customer's supplier qualification questionnaire. I have had this conversation with 40+ mid-market CEOs over the past 18 months, and the pattern is consistent: most knew the CSRD regulation existed, few had understood that the number their OEM customer was asking for required shop-floor data they did not have. This article is the working view of what energy KPIs actually are, what data model makes them defensible, and why 2026 is the year the Mittelstand's spreadsheet answer to this question stops being acceptable.

The regulatory landscape — three forces converging on the same shop-floor data

Three regulatory mechanisms are driving the urgency, and they need to be understood separately because they make different demands on the same underlying data:

The practical implication — and this is the point most mid-market CEOs have not fully internalised — is that most Mittelstand manufacturers are hit by force 4 (OEM supplier requirements) 12–24 months before force 1 (CSRD direct obligation) would catch them. A Tier-2 supplier in Baden-Württemberg with 180 employees is not in CSRD scope and never will be. But its three largest customers are, and the customer questionnaire arriving in Q2 2026 will ask for a kWh-per-part figure, a CO₂-per-part figure, and audit evidence behind both. The supplier that can answer with audit-ready data keeps the contract. The supplier that cannot, loses it — not in 2026, but in the next contract renewal cycle. This is the real transformation pressure, and it operates without the CSRD itself being formally invoked.

The Meter-Order Handshake — why the data problem is not a metering problem

When a customer asks for kWh per part, most plants' first instinct is to buy more meters. This is almost always the wrong instinct. The structural problem the pattern I call The Meter-Order Handshake addresses is not measurement density; it is measurement alignment. A production line running three products in a shift with a single energy meter at the main switch-board produces a perfectly accurate kWh-total figure for the shift. What it does not produce — and what cannot be reconstructed after the fact from that meter alone — is the allocation of those kWh to the three products individually, in the ratio that actually occurred. The data the customer needs is not kWh-total; it is kWh-per-SKU, and the difference between those two data points is an entire architectural discipline.

The handshake has three elements that must all be in place:

Plants that make the handshake work end up with less metering hardware than plants that do not — because the correct architectural answer to "how do I allocate plant-total energy to my 240 SKUs?" is never "install 240 meters". It is "install 8 well-placed meters and bind their streams to the MES order events." I have seen customers spend €200,000 on additional metering hardware that did not solve the allocation problem because the time-alignment problem was never addressed. I have seen other customers solve the problem for under €30,000 because the metering architecture was designed around the allocation question rather than around the meter catalogue.

The kWh-per-part calculation — five places where the arithmetic goes wrong

The canonical formula is trivial:

kWh per part = Energy consumed (kWh) / Parts produced (count)

The complications are not in the arithmetic; they are in the boundaries. Five specific places where what should be a simple calculation becomes materially wrong:

1. Good-part vs total-part denominator. The same plant can report kWh/part of 2.4 or kWh/good-part of 3.1 for the same period — a 29 % difference — purely because 22 % of output was scrap. The customer's PCF questionnaire almost always wants good-part denominator; the plant's internal dashboard almost always defaults to total-part. The mismatch between the two is what I call The Per-Good-Part Penalty — the visible cost, in energy terms, of quality losses that the OEE dashboard already should have been showing but that become concrete when expressed in kWh.
2. Idle and standby energy. A machine consumes energy during scheduled downtime, during unscheduled downtime, during changeover. Does this energy count against the parts produced before the downtime, after the downtime, or spread across the shift? The GHG Protocol says: allocate on a defensible basis and document the rule. Most plants do not have the rule written down, which means different shifts allocate differently, which means year-over-year comparisons become meaningless.
3. Shared-asset allocation. A compressor station supplies three lines. The hall HVAC runs continuously. The main chiller serves four processes. These shared-asset consumptions are real energy that real products consumed, but which products and in what ratio? Without an explicit allocation rule — proportional to kWh of the consuming line, to production count, to mass, to floor area — the kWh-per-part figure is indeterminate. This is what I call The Energy Allocation Gap, and it is the single most common reason CSRD auditors push back on submitted data.
4. Energy carrier boundaries. Electricity is easy. What about the natural gas used by the process oven? The compressed air treated as "free"? The steam raised in a central boilerhouse? A complete kWh-per-part figure converts all carriers to a common unit (kWh primary energy) using published energy-content factors, and a complete CO₂-per-part figure multiplies each by its own emission factor. Plants that report only electricity and call the resulting number "kWh per part" are materially understating — sometimes by a factor of two or more.
5. Time-averaging vs real-time. A monthly kWh/part of 2.4 averages over good days and bad days. The Monday-morning figure may be 3.1 because the line was still warming up and produced below rate; the Wednesday-midday figure may be 2.1 because the line was at steady state. Aggregation hides the variability that would otherwise point to specific improvement opportunities. ESRS E1 accepts monthly aggregates for reporting; operational improvement requires finer granularity.

CO₂ per part — Scope 2 dual reporting and the emission factor problem

The kWh calculation is the easy half. The conversion to CO₂ per part introduces two layers of complexity that almost every mid-market manufacturer underestimates. The first is the requirement, formalised in the GHG Protocol Scope 2 Guidance since 2015 but now enforced under ESRS E1, for what I call The Scope 2 Dual Report:

The consequence: every CO₂-per-part figure in a CSRD-compliant report must be reported twice, once under each methodology, with the difference explained. A plant with 100 % green-electricity contracts will show a market-based CO₂-per-part close to zero but a location-based figure at grid-average — both are true, both are required, and the spread between them is a legitimate discussion point in audit. Plants that report only one number are out of compliance regardless of which number they chose.

The second layer of complexity is the one I call The Emission Factor Decay. The German electricity grid's average CO₂ intensity has fallen from approximately 420 g/kWh in 2019 to approximately 380 g/kWh in 2024. That is a 10 % decline over five years, driven by renewable expansion, and the trend is continuing. The practical consequence: a plant that reports CO₂-per-part this year against a 380 g/kWh factor and last year against a 420 g/kWh factor can claim a CO₂-per-part improvement of 10 % while having changed absolutely nothing in its physical process. Year-over-year comparisons of CO₂-per-part without factor-normalisation are therefore meaningless, and auditors know this. ESRS E1 requires disclosure of the factor source and vintage for every reported figure; plants that cannot produce this disclosure get flagged.

The compressed air blind spot — and why it is almost always the first win

In every mid-market energy assessment I have been part of over the past decade, the same asset class has shown up as the largest untapped optimisation target: compressed air. The pattern I call The Compressed Air Blind Spot is the systematic under-measurement of the second-largest industrial energy sink after process heat. Three structural reasons:

• Compressed air is treated as free internally. No internal cost-accounting charges the user line for the compressed air it consumes. Consequently no one measures it.
• The efficiency losses are invisible. A compressor station delivering air to a line at 6 bar is running the compressor at the equivalent of 12× the electrical energy that the mechanical work at the end of the pneumatic line actually delivers. This 92 %-loss is the baseline for a well-maintained system; leaking systems routinely run above 95 %.
• Leaks are silent. A compressed-air leak costs real money in electricity consumed but makes no operational noise that anyone cares about. The industry rule of thumb is that 20–30 % of compressed-air energy is lost to leaks in un-managed systems, and I have personally seen plants where the figure was above 40 %.

The operational consequence for energy KPIs: a plant that adds compressed air to its kWh-per-part calculation for the first time typically sees the figure jump by 25–40 %, and the subsequent search for leaks typically finds three to five obvious ones in the first week and ten to twenty more over the next three months. This is not a theoretical optimisation opportunity. It is a line item that, in my experience, has paid for the entire energy-KPI implementation project in most mid-market customer cases inside six months.

Industry-specific energy-intensity benchmarks

The kWh-per-part figure is only useful in context. Absolute values vary over three orders of magnitude depending on process type, which is why cross-industry benchmarking is almost always misleading. The ranges I have observed across the SYMESTIC customer base, validated against public literature where available:

The operational discipline: benchmark against the lower bound of your process class, not against the industry average. The 25th-percentile plant in any process class is typically 30–40 % more energy-efficient than the 50th-percentile plant, and the gap is almost entirely explained by three variables — idle-state management, compressed-air discipline, and the quality of the metering-to-order handshake that lets the plant see what it is doing.

The CSRD Audit Bridge — connecting the shop floor to the sustainability report

The positive pattern that makes everything above work in practice is what I call The CSRD Audit Bridge: the architected data path from individual meter reads and production-order events on the shop floor, through the MES, into an aggregation layer that produces audit-ready Scope 2 and product-carbon-footprint figures in the format the sustainability report requires. Seven components:

The bridge is the architectural output of the energy-KPI programme. Plants that have it in place typically handle a customer PCF request in under two hours. Plants that do not handle the same request in two to six weeks, involve four to six people, and produce a number they cannot defend if the customer's auditor pushes back. The productivity difference is the difference between treating the kWh-per-part question as a shop-floor data problem, which it is, and treating it as a spreadsheet problem, which it never was.

From a mid-market automotive Tier-2 supplier in Baden-Württemberg, autumn 2024: The customer was a family-owned supplier of precision metal components — stamping, deep-drawing, surface treatment — serving three of the major German OEMs directly and two Tier-1s indirectly. Roughly 280 employees, €62 million turnover, ISO 50001 certified since 2019, IATF 16949 since forever. I had known the CEO for about six years through a Mittelstand industry roundtable. He called me in September 2024 because the commercial director had received, on the same day, a CSRD supplier-data questionnaire from one OEM and a PCF request from a Tier-1 for a specific family of brake-system brackets. Both asked for kWh per part and CO₂ per part per SKU, location-based and market-based, with methodology disclosure, for the fiscal year 2024. The internal response had been, he said with the kind of smile German CEOs use when something is serious, "we have ISO 50001, this should be straightforward." Two weeks of internal work later, it was not straightforward. The plant had five main-hall electricity sub-meters, three gas meters, one compressed-air flow meter at the compressor station, and a well-maintained ISO 50001 energy-management system that produced monthly plant-total consumption figures against plant-total output. What it did not produce was per-SKU allocation. The 240 active SKUs ran on six lines. The lines shared the compressor station, the chiller, the hall HVAC, and a central wastewater treatment facility. The monthly energy manager's spreadsheet allocated everything proportional to mass of output — which was defensible but not defensibly per-SKU, because two SKUs of the same mass could consume materially different energy depending on which press had been used, how many tooling changes had occurred, and whether the surface-treatment line had been in use. When I arrived on site for the first working session, the energy manager walked me through three weeks of attempted Excel reconstruction. It was careful, thoughtful work. It was also, at the SKU level, impossible, because the per-order energy consumption had never been recorded — only the per-month plant totals. We built the remediation in three phases over the following four months. Phase 1: time-alignment of existing meters. The five main sub-meters and the compressed-air meter were already recording at 15-minute intervals but had clock drift of up to 90 seconds against the MES. We brought them onto a common NTP server, installed the SYMESTIC connector to pull the streams in real-time, and backfilled three weeks of aligned data for validation. Phase 2: allocation-rule definition. We convened the energy manager, the production planner, the quality director, and the commercial director and wrote — literally wrote, on a whiteboard, then in a signed rule document — the allocation rules for each shared asset. Compressor: allocated proportional to the compressed-air flow measured downstream of each of the six line-inlet manifolds, requiring one additional flow meter per line (installed within three weeks). Hall HVAC: flat allocation across lines proportional to floor area. Chiller: proportional to chilled-water flow per process, which required two additional meters. Wastewater: allocated by water consumption, already metered. The rule document ran to eleven pages and was signed by all four participants; it became the audit-defence baseline. Phase 3: factor layer and dual reporting. We pulled the Umweltbundesamt German grid factor for 2024 as the location-based reference, confirmed with the utility provider that the plant's existing green-tariff contract had full AIB guarantees-of-origin coverage for the year, and set up the market-based reporting pathway at near-zero g/kWh. The PCF aggregation layer produced location-based and market-based numbers side by side for every SKU. In the process of building the allocation layer, we discovered — and this was the nebenbefund, the side-finding that paid for the entire project — that the compressor station was consuming 14 % of the plant's total electricity, substantially above the 8–10 % benchmark for a stamping operation of this size. An ultrasonic leak survey over one weekend found 23 significant leaks, the largest on a dead-end spur that had been feeding a scrapped machine for roughly four years. Closing the leaks and correcting the compressor control strategy brought compressed-air energy down by 31 % over the following quarter. The customer's return on the full project, including the meter additions, the MES integration work, and the leak remediation, was positive in month seven. The OEM CSRD response was filed in January 2025, with location-based and market-based per-SKU figures, allocation-rule disclosure, and factor provenance. The auditor had two clarifying questions on allocation; both were answered with the signed rule document. The questionnaire was accepted without revisions. The deeper lesson I have repeated at every Mittelstand customer since: the CSRD kWh-per-part question is not, fundamentally, a sustainability question. It is a shop-floor data-architecture question that the sustainability director is asking on behalf of the CFO, who is being asked by the OEM commercial director, who has been asked by his auditor. Plants that treat it as a sustainability problem to be solved with consultants and spreadsheets take six to twelve months and produce figures they cannot defend. Plants that treat it as what it is — a meter-to-order data-linkage problem, which any competent MES already has most of the infrastructure for — solve it in one quarter and come out with an operational asset rather than an annual compliance burden. I have watched the two approaches run in parallel at two comparable companies. The first is still in its CSRD project two years later. The second was asked last autumn to present its approach at a regional industry association meeting. The difference was not budget, not team, not strategic commitment. The difference was whether the problem was framed as a spreadsheet problem or a data-architecture problem. Framing decides outcome.

The seven disciplines of defensible energy KPIs

FAQ

What are energy KPIs in manufacturing?
Energy KPIs translate a production process's energy consumption and associated emissions into numbers that can be compared, reported, and audited. The two foundational metrics are kWh per part (energy consumed per manufactured unit) and CO₂ per part (emissions caused per manufactured unit, typically split into location-based and market-based under GHG Protocol). Their per-good-part variants and their aggregations to SKU, line, shift, and plant are the operational working set.

What is the difference between kWh per part and kWh per good part?
kWh per part divides total energy consumed by all produced units including scrap. kWh per good part divides by usable units only, which makes the energy cost of quality losses visible. The difference between the two — what I call The Per-Good-Part Penalty — is often 10–30 % at mid-market plants with typical scrap rates, and CSRD auditors and OEM customers almost always want the good-part denominator for product carbon footprint reporting.

What is The Scope 2 Dual Report?
Under GHG Protocol Scope 2 Guidance and ESRS E1, every electricity-related CO₂ figure must be reported under two methodologies: location-based (grid-average emission factor for the geographic region) and market-based (supplier-specific factor reflecting actual procurement contracts and green-electricity guarantees). Plants that report only one number are out of compliance regardless of which methodology they chose.

Does my company have to comply with CSRD?
Directly, only if you are in Wave 1 (EU-listed, already in scope) or Wave 2 (large non-listed EU entities meeting two of three thresholds: 250+ employees, €50M+ turnover, €25M+ balance sheet — first reports for fiscal year 2025, filed in 2026). Indirectly, via Scope 3 supplier requests, every manufacturer supplying a Wave 1 or Wave 2 reporter faces the same data demands one to two years earlier than the direct obligation would hit. In 2026, the indirect pressure via OEM supplier questionnaires is the dominant practical force on the Mittelstand.

What is The Energy Allocation Gap?
The inability to allocate shared-asset energy consumption (compressor station, central chiller, hall HVAC, wastewater treatment) to individual products in a defensible ratio. Without explicit written allocation rules per shared asset, per-SKU kWh and CO₂ figures are indeterminate, and ESRS E1 auditors systematically reject submissions that cannot document the rule. The fix is an architectural one: written allocation rules, additional flow or sub-metering where proportional allocation is required, and the rule document signed by cross-functional stakeholders for audit defence.

Can energy KPIs be calculated in Excel?
In principle, for a single line with a single product and a single meter, yes. In practice, for any plant with multiple products, shared assets, dual Scope 2 reporting, and customer-PCF requests, Excel reconstruction after the fact cannot produce per-SKU per-order figures because the underlying per-event data was never captured. The 2026 Mittelstand test: can you answer a customer PCF questionnaire for a specific SKU in two hours, or does it take two weeks? Spreadsheets produce the second answer.

What is The Compressed Air Blind Spot?
The systematic under-measurement of compressed air in industrial energy accounting. Compressed air is the second-largest industrial energy sink after process heat, typically 8–15 % of plant electricity at a well-managed mid-market site, and over 20 % at un-managed sites because leaks routinely waste 20–30 % of compressor output. Most plants do not meter compressed air beyond the main compressor station, so per-line and per-product allocation is impossible. Adding compressed-air metering is almost always the single highest-ROI first step in an energy-KPI programme; in my experience, it pays for the full programme within six months in the majority of Mittelstand cases.

How does energy KPI tracking relate to ISO 50001?
ISO 50001 is the Energy Management System standard. It defines the management discipline — policy, planning, operational control, checking, management review — but it does not prescribe a specific measurement architecture. A plant can be ISO 50001 certified without having the per-SKU allocation capability that CSRD and OEM PCF requests require. In 2026, ISO 50001 is necessary but not sufficient; the plant-level management discipline has to be extended with the product-level measurement architecture to meet customer demands.

What is The Emission Factor Decay?
The annual decline of grid electricity CO₂ intensity as renewable capacity expands. The German grid average fell from ~420 g/kWh in 2019 to ~380 g/kWh in 2024 — a 10 % decline with no change in industrial behaviour. Year-over-year CO₂-per-part comparisons without factor-vintage normalisation therefore contain a systematic drift that auditors flag. ESRS E1 requires factor source and vintage disclosure on every reported figure; plants that cannot produce this disclosure fail the compliance test.

Why does 2026 matter specifically?
Three mechanisms converge. Wave 2 CSRD reporters file their first fiscal-year 2025 reports. CBAM moves from transitional to definitive regime on 1 January 2026. OEM supplier questionnaires, which started appearing on RFQs in 2023–2024, become mandatory for existing business contracts at renewal. A Mittelstand manufacturer that enters 2026 without the data architecture to respond is not noncompliant with a regulation — it is noncompliant with the contractual expectations of its largest customers. That is a harder problem to recover from than a regulatory finding.

Related: MES: definition, functions & benefits · OEE · Peak shaving · Energy monitoring · ISO 50001 · CSRD reporting · Scope 2 emissions · Product carbon footprint · CBAM — Carbon Border Adjustment Mechanism · Real-time production data · Machine data integration · Production data acquisition · Audit trail · End-to-end traceability · Industrial data historian · Schedule adherence · Composable MES · Production metrics · Process data · For metal processing · For automotive suppliers · For food & beverage manufacturers · For COOs & plant managers · For operational excellence. External references: EFRAG — ESRS E1 Climate Change · CSRD Directive (EU 2022/2464) · GHG Protocol Scope 2 Guidance · European Commission — CBAM · ISO 50001 — Energy management · Umweltbundesamt — German grid emission factors.

About the author

Uwe Kobbert

Founder and CEO of SYMESTIC GmbH. More than 30 years in the manufacturing industry. Dipl.-Ing. Nachrichtentechnik/Elektronik. Career path: Consultant at SAS 1989–1992 (industrial consulting, Heidelberg), Division Manager Industry at STERIA Software Partner 1992–1995 (process control systems and manufacturing execution systems for food and beverage industry, including early energy-monitoring modules). Founded SYMESTIC GmbH in Dossenheim in 1995 and has led the company since. Built SYMESTIC from classical on-premise MES projects through the strategic re-architecture to cloud-native in the mid-2010s, resulting in the SYMESTIC Cloud MES Platform deployed in 18 countries across four continents with over 15,000 connected machines. Nominated for the Großer Preis des Mittelstandes. Self-funded, no external investors. Expertise: Manufacturing Execution Systems, OEE and production KPIs, energy monitoring and CSRD-compliant reporting, shopfloor management, cloud-native manufacturing software, Industry 4.0, Lean Production, industrial automation, process control systems, ERP-MES integration, PLC programming, JIT/JIS processes, batch manufacturing, automotive production, food industry. Primary perspective on the CSRD energy-reporting challenge: it is a shop-floor data-architecture problem, not a sustainability problem, and framing decides outcome. · LinkedIn