Somewhere in your plant right now, a sustainability manager is building a spreadsheet that pulls utility bills, production volumes, and emission factors into a monthly carbon number for corporate reporting. It’s slow, it’s manual, and it’s about to become legally insufficient. CSRD’s phase-in schedule and CBAM’s move from transitional to definitive reporting are both landing squarely on manufacturing operations in 2026, and the granularity regulators and auditors expect — product-level, site-level, sometimes shift-level carbon intensity — simply cannot come from a spreadsheet reconciled once a month. It has to come from the same systems that already track your production data. That means MES.
This is not a sustainability-office problem anymore. It’s an ISA-95 modeling problem, a data-acquisition problem, and an equipment-hierarchy problem — which is to say, it’s your problem.
Why this lands on MES and not just the ESG team
CSRD requires reported metrics to be auditable with the same rigor as financial data — traceable to source, defensible under assurance review, not reconstructed after the fact from invoices and estimates. CBAM, for goods crossing into the EU, requires embedded emissions calculated per tonne of product using actual production data where available, not just default values, once the definitive regime phases in. Auditors and customs authorities are going to ask the same question a plant manager asks about a scrap rate: where did this number come from, and can you show me the record?
Your MES already answers that question for OEE. It timestamps downtime, ties it to a work order and a piece of equipment, and reconciles it against a shift schedule. Carbon intensity needs the exact same treatment: a number tied to a specific asset, a specific production run, and a specific time window, captured automatically rather than typed in after the fact. If you’re serious about hitting 2026 deadlines without a headcount explosion in the sustainability office, the fastest path is to make energy and carbon a native MES data type, not a separate system that reconciles against MES after the month closes.
Where the data actually lives — and where it doesn’t yet
The good news is that most plants already have more of this data than they think. Modern PLCs and drives frequently expose power consumption tags already — VFDs report kW draw, compressors and chillers often have built-in energy monitoring, and many utility submeters installed for demand-charge management sit on the same network as your production equipment, just never wired into MES. The unglamorous first step of any energy-intensity project is an inventory: walk the equipment hierarchy you already have in your MES or historian and ask, for each asset, whether an energy tag exists somewhere in the automation layer, in a building management system, or on a standalone meter with a Modbus or BACnet interface.
Where you’ll find real gaps is at the process-line level for older equipment — legacy machines with no energy instrumentation at all, or shared utilities (compressed air, steam, chilled water) metered at a header level but consumed by a dozen work centers downstream. For CSRD and CBAM purposes, you don’t necessarily need to instrument every motor. You need enough metering resolution to allocate energy and emissions to a product or process step defensibly, and allocation methodology (submetering versus engineering estimates versus production-based apportionment) is something your auditors will scrutinize as much as the raw numbers. Where new sensors are genuinely required — shared utility lines with no submetering, or CBAM-in-scope processes with no existing instrumentation — budget for it as a metering project with its own commissioning and validation, not a footnote on the MES rollout.
Model it in the hierarchy you already have
Here’s where teams go wrong: they treat carbon reporting as a reason to redo the ISA-95 equipment hierarchy from scratch. Don’t. Your hierarchy — enterprise, site, area, work center, unit — already represents the physical and organizational structure production data rolls up through. Energy and emissions should roll up through the exact same tree.
- Attach energy tags at the level they’re actually metered — often work center or area, sometimes only site — and let the hierarchy handle apportionment downward rather than inventing a parallel structure.
- Treat emission factors as a reference data set, not a hardcoded constant. Grid emission factors change by region and year; if you’re consuming purchased electricity versus on-site generation versus increasingly common power purchase agreements, the factor differs and needs to be versioned and auditable, the same way you version a BOM or a recipe.
- Bind energy consumption to the same work-order and batch context OEE already uses. If your MES already ties a downtime event to a work order, tie a kWh reading to that same work order. You get product-level and even batch-level carbon intensity almost for free, because the context-linking work is already done.
This is also where CBAM gets specific in a way CSRD doesn’t: CBAM wants embedded emissions per tonne of covered goods (steel, aluminum, cement, fertilizers, hydrogen, and a few other categories), calculated from actual production data. If any of your product lines fall into a CBAM-covered category, that product-level binding isn’t a nice-to-have. It’s the whole point.
Make carbon the sixth loss category
OEE works as a management tool because it decomposes a single confusing number into three legible categories — availability, performance, quality — each with its own root causes and its own owner on the floor. Carbon intensity should get the same treatment, not bolted on as a separate KPI living in a different dashboard that nobody on the shop floor ever opens.
Practically, that means defining kWh/unit or kg-CO2/kg-product as a tracked variable at the same work-order and shift resolution as OEE, and asking the same diagnostic questions: is intensity worse during changeovers and startup (an availability-adjacent loss), worse during degraded-rate running (a performance-adjacent loss), or worse because of rework and scrap consuming energy without producing shippable product (a quality-adjacent loss)? Framed that way, energy stops being an abstract sustainability metric and becomes something a line supervisor can actually act on — because idle-but-powered equipment, slow cycles, and scrap are exactly the failure modes OEE already trains people to hunt down. You’re not asking operators to care about a new thing; you’re asking them to notice that the thing they already fight shows up on a second scoreboard too.
This framing also happens to produce the audit trail regulators want, as a side effect rather than a separate initiative. A number generated by the same system, same context-binding, and same historian that already produces your quality and downtime records inherits that system’s traceability. A number copied from a utility bill into a spreadsheet inherits nothing.
What to actually do in the next few quarters
Don’t wait for a perfect enterprise sustainability platform decision before starting. Start by auditing which of your existing PLC, drive, and meter tags already expose energy data and get them into the historian you’re already using for production data. Extend your ISA-95 equipment model with energy attributes at the levels where metering genuinely exists, rather than guessing at finer granularity than your instrumentation supports. Version your emission factors as reference data with change history. And put kWh and carbon intensity on the same review cadence as OEE — daily or shift-level on the floor, rolled up monthly for the sustainability office — so it’s a living operational metric long before an auditor ever asks to see it.
The plants that treat this as an MES extension will have auditable numbers by the time CSRD and CBAM deadlines bite. The plants that treat it as a reporting exercise for the sustainability office will be reconciling spreadsheets against production records at the worst possible time, with an auditor watching.
This article was written with the assistance of artificial intelligence. While we aim for accuracy, the information may be incomplete, out of date, or incorrect, and should be independently verified before you rely on it for any decision. It is provided for general information only and does not constitute professional advice.
