IED 2.0: The Meters You Need Before July 2027

Walk the utility corridor of a large food plant and you'll find the meter. Brass body, wall bracket, a glass window with a rolling counter turning behind it. It has measured the incoming water main to within a couple of percent for a decade, faithfully, without complaint. Now look at the terminal block on its flank. In plenty of plants the pulse-output wires are still coiled in the loom, cut back, taped off, landed on nothing.

The meter works. Its reading goes onto a clipboard once a month, gets typed into a spreadsheet, and reconciles against the utility invoice near enough. Nobody is cutting corners. But that plant can't produce, on demand, a defensible figure for cubic metres of water per tonne of product across a calendar year, resolved by process, with the raw series still sitting behind it. Inside two years, it will have to.

The revised Industrial Emissions Directive passed its Member State transposition deadline on 1 July 2026. That date belongs to national ministries. The one that belongs to operators falls a year later: by 1 July 2027, every installation already inside the directive's scope must have an environmental management system prepared, implemented, and audited for the first time by an accredited conformity assessment body or a licensed environmental verifier, then re-audited at least every three years. The Commission puts the directive's reach at roughly 37,000 industrial installations. What changed isn't the stack limit. It's that resource use (water, energy, materials, waste) stops being something a plant merely consumes and becomes something a plant has to measure, defend, and publish.

Emissions have a quality-assurance regime. Water doesn't.

Consider how much apparatus already stands behind a single stack number. A waste-to-energy plant measures dust, HCl, SO2, NOx and CO continuously, on an automated measuring system permanently installed on site, and the BAT conclusions for waste incineration name the EN standards that keep it honest: EN 15267 for certifying the instrument, EN 14181 for its ongoing quality assurance. Calibration function. Drift check. Annual surveillance test. A half-hourly average counts as valid only if the analyser wasn't in maintenance or malfunction while it was recording. Decades of institutional practice sit underneath that one figure.

Now look again at the water meter in the corridor. No certification scheme. No drift check. No definition of a valid averaging period, because nobody ever needed one. It was installed to settle an invoice, and that's the whole of what's ever been asked of it.

Article 14a closes that gap, and it repays reading in the original rather than in summary. The environmental management system it mandates is not a binder on a shelf. It has to carry environmental policy objectives that include preventing waste, optimising resource and energy use and water reuse, and reducing the use or emission of hazardous substances. It has to carry objectives and performance indicators for the installation's significant environmental aspects, taking account of the benchmarks set out in the relevant BAT conclusions. It has to carry a chemicals inventory, with a risk assessment and an analysis of whether safer substitutes exist. Where a company already sits under the energy-audit duty, the EMS has to absorb the results of that audit and the measures taken on its recommendations.

That last hook repays following, because it's already live. Under the recast Energy Efficiency Directive, an enterprise averaging more than 85 TJ of energy a year over the previous three years (about 24 GWh, all carriers taken together) must run an energy management system certified by an independent body, in place by 11 October 2027. For energy management the international standard is ISO 50001. Below that line, above 10 TJ, an energy audit falls due every four years instead. So for a good many sites the measurement duty didn't arrive with the revised directive at all. It arrived earlier, from a different direction, and the EMS is merely where it now has to be shown.

Binding for water, indicative for everything else

Here the summaries and the statute part company, and the gap caught us out on first reading. The Commission's own page on the revised directive says permits must contain "binding quantitative resource efficiency requirements for materials, water and energy, as appropriate". Open Article 15(4) and the architecture turns out to be layered. The competent authority sets binding ranges for environmental performance, as laid down in the BAT conclusions. On top of those it sets environmental performance limit values concerning water, and those are binding. For waste, and for resources other than water, which is to say energy and materials, it sets indicative environmental performance levels.

Water gets a limit value. Energy and materials get a level, and the word the legislator chose is indicative.

So does indicative mean the wires can stay taped off? It does not. The indicative level is still set in the permit. It still sits inside a binding range drawn from the BAT conclusions. The benchmark still goes into the EMS, the EMS still gets audited every three years by an accredited body, and the relevant information in it still has to be made available on the internet, free of charge, without even a registration wall to slow a reader down. An auditor asking how you arrived at your specific energy consumption is not asking a gentler question because the number carries the word indicative in front of it.

Indicative is a legal word, not an engineering one. It changes what a regulator can fine you for. It changes nothing about what you have to measure.

The instrumentation bill comes out identical either way. Which is why the binding-versus-indicative argument, which consumed a great deal of consultation energy, turns out to be close to irrelevant for the person who has to specify the meters.

The denominator is the part that breaks

Resource performance in the BAT conclusions is never an absolute quantity. It's a ratio, and more precisely it's a yearly average of a ratio, which is a different and considerably harder thing to produce. The BAT conclusions for the food, drink and milk industries set out the arithmetic plainly: specific energy consumption is the final energy consumption, in MWh per year, of the specific processes concerned, divided by the activity rate in tonnes or hectolitres per year. Packaging doesn't count toward the weight of the product.

Read that definition twice. Two problems fall out of it.

The first is the boundary on the numerator. Not the site. Not the incomer. The specific processes concerned. A dairy's incoming HV supply tells you what the whole site drew, office block and car-park lighting and effluent-plant blowers included. It says nothing about what pasteurisation drew. Closing that distance means sub-metering along the process boundary the BAT conclusions actually name, which in most plants means current transformers on motor-control-centre outgoers and heat meters on the steam and hot-water headers, rather than one clamp at the substation. Steam is where the exercise usually founders. Steam gets distributed, condensed, flashed and returned, and a plant that meters its gas but never its headers can tell you exactly what the boiler burned while knowing almost nothing about what the process used.

The second problem is that the denominator lives in a different building. Activity rate is production: tonnes, or hectolitres. That figure comes out of the MES or the ERP, up at the top of the ISA-95 stack, and it typically arrives as a shift roll-up or a monthly close. The energy figure comes from the historian, at fifteen-minute or one-second resolution, on a clock disciplined by NTP if somebody thought to set it. To compute the compliance number you have to join them. Two systems, two owners, two timebases, one ratio.

And the sectors don't even agree on what to divide by. Brewing is measured per hectolitre of product. Dairies are measured per tonne of raw material, which means milk intake at the tanker bay, not yoghurt out the door. Those are different meters, in different rooms, owned by different departments, and a plant that instruments the wrong one has bought a number it can't use.

Cleaning-in-place is the quiet one. A CIP skid runs to a schedule of its own, draws water and heat in bursts that bear no relation to the production rate at that moment, and it usually hides behind a header nobody sub-meters. A plant that improves its specific water consumption by tightening rinse cycles has done real engineering work. Proving it did so, across a year, against a denominator collected by a system it doesn't talk to, is a data problem long before it's an engineering one.

Where this breaks down, in practice, is at the join. The common mistake is to treat the yearly average as forgiving, on the theory that a year is long and errors wash out. They don't wash out. A missing fortnight of production data, a meter swapped mid-year without a note in the log, a long batch that straddles a month boundary and gets counted at both ends: none of it announces itself in an annual figure. It surfaces instead as a plausible number that happens to be wrong, which is the worst kind there is, because nobody catches a plausible number until an auditor asks to see the workings behind it.

Six power terms and a fuel nobody meters

Waste-to-energy makes the derived-number problem impossible to look away from. Energy performance on an incineration line isn't a meter reading at all. It's an efficiency, expressed as a percentage, and per the waste-incineration BAT conclusions it gets assembled out of six separate power terms. The electrical power generated. The thermal power supplied to the heat exchangers. The directly exported thermal power, less the thermal power of the return flow. The thermal power produced by the boiler. The thermal power used internally, for duties such as flue-gas reheating. And the thermal input to the furnace, expressed as lower heating value.

Each of those is an instrument, or a small committee of them. The export term alone, exported thermal power minus the return flow, quietly promotes a district-heating return-temperature sensor into a compliance instrument. Nobody at the plant thinks of it that way. It was installed for heat billing, and it is probably calibrated on that schedule.

Then there's the last term, and it's the one that ought to keep an instrumentation engineer awake. Thermal input, expressed as lower heating value. No inline calorimeter rides on a grab crane. The heating value of mixed municipal waste swings bin to bin and season to season, wet Monday against dry Thursday, and no single instrument measures it. It is inferred, in practice from a heat balance drawn around the boiler, which means the denominator of the compliance ratio gets computed from the plant's own steam measurements. Per the same BAT conclusions, a plant is judged against a gross electrical efficiency of 25-35 % when new and 20-35 % when existing, or a gross energy efficiency of 72-91 % where it delivers heat. That verdict inherits every bias in every flow meter used to derive it, along with whatever blowdown went unaccounted for.

Nobody runs a QAL2 on a steam flow meter. And that asymmetry is worth sitting with, because it's the shape of the whole problem: on the emissions side, a mature standards regime governs how the number is measured. On the resource side, the number is frequently a calculation, and the calculation answers to no regime whatsoever.

Your kilowatt-hours per tonne, on a public website

Pressure won't arrive mainly through the limit value. It'll arrive through disclosure, and that mechanism is already law. Regulation (EU) 2024/1244 establishes the Industrial Emissions Portal, which the Commission runs with the European Environment Agency, and repeals the old pollutant register with effect from 1 January 2028, the date the new rules start to apply. The Portal carries pollutant releases and off-site waste transfers, as its predecessor did. It also carries something genuinely new: data on the use of water, energy and relevant raw materials.

Alongside that, contextual information. The regulation's own words are production volume and number of operating hours.

Numerator and denominator. Published, per installation, for every large industrial site in the Union. Anyone with a spreadsheet will be able to compute your specific energy consumption and rank it against every comparable plant in your sector, whether or not the indicative level in your permit was ever crossed. The relevant parts of your EMS go on the internet as well, under Article 14a. The benchmark you set for yourself: on the internet. Your performance against it: on the internet.

For operators who've already done the metering, that's a shop window. For the ones who haven't, it's an audit with an unlimited number of auditors. The directive does hand Member States a penalty floor for the worst cases, since administrative fines for the most serious infringements by a legal person must reach a maximum of at least 3 % of annual Union turnover. But the fine isn't the interesting risk. Being the visibly worst dairy in your Member State, on a public dashboard, with a retail buyer's sustainability team reading it, is exposure of an entirely different character.

Timelines then diverge by sector, and it pays to know which queue you're standing in. Metals producers sit in Annex I point 2, which puts them in the first wave for transformation plans under Article 27d: an indicative plan inside the EMS by 30 June 2030, setting out how the installation reaches a clean, circular and climate-neutral footing. Food and drink plants and waste incinerators aren't on that first list. Their transformation-plan duty comes later, attached to the permit reconsideration that follows their next BAT conclusions published after 1 January 2030. The sectors the revision newly pulled into scope, metal mining and large-scale battery production among them, run on a longer clock still. Not every site is on the same schedule. Every plant already in scope is on the 2027 one.

What to wire before the auditor comes

None of this calls for a moonshot. It calls for the boring layer, built deliberately and early, and five things are worth getting right.

• Sub-meter to the boundary the BAT conclusions name, not to the boundary your electricity supplier cares about. If the sector's indicator runs per tonne of raw material, the metering has to resolve the processes that consume that raw material, and not the site as one lump.
• Put production on the same clock as the utilities. A tonnage that arrives as a monthly total cannot be divided into a fifteen-minute energy series in any manner an auditor will respect. Getting production counts into the historian, timestamped where they're generated, is usually the highest-value piece of plumbing on the whole list.
• Keep the raw series, not the roll-up. When somebody asks how a figure was derived, an average of twelve invoices is not an answer. Retention of the underlying data is the thing that converts a number into a defensible number.
• Version the calculation. Specific consumption is a formula with a boundary, and boundaries move: a new sub-meter here, a re-piped header there, a product mix that shifts what counts as raw material. Write the definition down, record the date it changed, keep both.
• Treat derived values as instruments in their own right. A lower heating value, a boiler efficiency, a return-flow delta: if a compliance figure leans on it, it needs an owner, a stated uncertainty, and a review cadence, exactly as a transmitter would.

Most of that is telemetry and data engineering rather than process engineering, which is precisely why it falls between departments and stays there. Instrumentation owns the meter. IT owns the historian. Operations owns the tonnage. Compliance owns the report, and tends to discover in the fourth quarter that the three groups feeding it never agreed on a boundary in the first place. Stitching those together, from the sensor through to an audit trail somebody can defend, is the work we do on our edge telemetry and analytics platform, and the honest scope of it, at most sites, is counted in months rather than weeks.

Go back to the corridor. The brass meter is still there, still accurate, still turning over. The two pulse wires are still taped off in the loom, because for a decade nothing on earth depended on them. Something does now, and it has a date on it.