Counting the Kilowatts: Why ESOS Phase 4 Pushes Factories to Sub-Meter the Shop Floor

What does ESOS Phase 4 ask a factory to measure?

ESOS Phase 4 asks a qualifying factory to account for at least 95% of its total energy consumption, to calculate at least one energy intensity ratio for each organisational purpose where it uses energy (buildings, transport, industrial processes and anything else), and to use verifiable data wherever that is reasonably practicable. The Environment Agency's Phase 4 guidance, published on 30 July 2026, sets the dates: an organisation qualifies if it is a large undertaking on 31 December 2026, and the notification of compliance is due by 5 December 2027.

A large undertaking, for ESOS purposes, is one that employs 250 or more people, or has annual turnover above £44 million and a balance sheet total above £38 million. That catches a great many UK manufacturers who would not describe themselves as large: a food processor with two sites in the Midlands, a plastics moulder supplying the automotive trade, a contract packer running three shifts.

Phase 4 also carries forward the obligations introduced in Phase 3 and adds a few more. The action plan is due by 5 December 2028, followed by annual progress updates in 2029, 2030 and, new for this phase, a third update by 5 December 2031. For measures that achieved savings during the compliance period, the organisation must now identify the energy saving category and the kWh saved. The Environment Agency will also publish each organisation's total and significant energy consumption broken down by purpose. Taken together, these changes mean energy figures that used to sit in a consultant's report will now be published, tracked year on year and compared against what the organisation said it would do.

Why does factory energy data stop at the main meter?

Most UK factories measure energy in exactly one place for each fuel: the half-hourly electricity incomer the supplier bills from, and the gas meter at the boundary. Those meters are accurate and their data is verifiable, which suits ESOS well for the 95% coverage test. They say nothing about where the energy went once it crossed the site boundary.

Inside the fence, a typical plant has a compressed air system, process heating (ovens, dryers, moulding machines, wash systems), refrigeration if it handles food, HVAC for offices and clean areas, lighting, and the production machines themselves. The Carbon Trust's guidance puts compressed air alone at around a tenth of the electricity used by UK industry, and it is the usual first suspect in any energy audit because leaks and running compressors over the weekend are common and cheap to fix. Without a meter on the compressor house, nobody can say whether a leak campaign saved 40,000 kWh or 4,000.

This is where ESOS auditors earn their fee and where the trade-offs start to bite. The guidance allows reasonable estimates when verifiable data is not available, provided the method is recorded in the evidence pack. Estimates from nameplate ratings and run hours are acceptable for scoping. They are a weak basis for an energy intensity ratio per production line, and weaker still for a progress update in 2030 claiming that a compressor replacement saved a specific number of kWh. An estimate built on assumptions in 2027 and compared with another estimate in 2030 measures the auditor's assumptions, and the savings it reports will be hard to defend if the Environment Agency asks.

How much sub-metering does an ESOS audit actually need?

An ESOS audit needs enough sub-metering to separate the significant energy uses that the action plan will target, and no more. ESOS does not mandate sub-metering, and we would discourage anyone from metering every distribution board for the sake of completeness. The useful question is narrower: which three to six loads account for most of the site's consumption, and which of them are likely to appear in the action plan?

For most manufacturers the answer is a short list:

• Compressed air: the compressor house, metered at the incoming supply to each compressor, ideally alongside a flow meter on the main header.
• Process heat: ovens, furnaces, moulding machines or wash systems, often the largest single electrical or gas load on site.
• Refrigeration and chilled water: the plant room for food and drink, pharmaceutical and some plastics sites.
• HVAC and building services: offices, welfare and controlled-environment areas, needed for the buildings purpose ratio.
• Each production line or hall: so that kWh per tonne, per unit or per pack can be calculated where the product is made.

Five to twenty meters on a single site is typical. At that scale the meters are rarely the expensive part: modern clamp-on current transformer meters with a Modbus or pulse output cost a few hundred pounds each. The installation and the data path back to somewhere useful is where the money goes.

What are the options for getting sub-meter data off the factory floor?

There are four practical ways to get sub-meter readings back to an energy monitoring system, and each suits a different site.

Hard-wired Modbus or Ethernet is the default for new builds and refurbished switch rooms. It is reliable and has no ongoing radio cost, but on a brownfield site each meter needs a cable run from a plant room to the nearest network point. Through fire-stopped walls, across a live production hall and up into trunking that is already full, a single run can cost more than the meter and take a shutdown to install. Five meters is manageable, while twenty across three buildings usually is not.

LoRaWAN is the low-cost wireless option and is well suited to energy data: a 15-minute kWh reading is a few bytes, battery life runs to years, and a single gateway covers a large site. Its limits appear when you want more than kWh. Low data rates rule out waveform capture, power quality analysis or the vibration data a maintenance team might later want from the same motor.

Public cellular (NB-IoT or LTE-M SIMs in each meter gateway) avoids building any network, but plant rooms, steel-clad halls and basement switch rooms are where public mobile signal is weakest. A meter that drops off the network for a week in February leaves a gap in the reference period, and the gap has to be filled with an estimate.

Factory Wi-Fi is usually present and usually the wrong home for meter data. Corporate Wi-Fi is designed around laptops and handhelds, its access points are rarely placed with switch rooms in mind, and IT teams are right to be wary of attaching dozens of headless OT devices to the corporate network.

When does private 5G make sense for an energy monitoring system?

Private 5G makes sense for energy monitoring when the factory has, or is planning, other shop-floor uses for the same network. On its own, a 15-minute kWh reading does not need 5G; LoRaWAN will carry it more cheaply. The case changes when the energy data is the first workload on a network that will also carry condition monitoring on the motors, AMR traffic, vision inspection or handheld scanners.

A licensed private network in the Ofcom shared access bands (n77, 3.8 to 4.2 GHz) gives the plant a single, managed radio layer that reaches switch rooms and compressor houses as reliably as the warehouse floor. The meter gateways sit on their own network slice or VLAN, isolated from corporate IT, which answers the OT security question that blocks the Wi-Fi route. Coverage is planned and measured with the plant engineer, so the dropout risk that undermines public cellular is designed out.

There is also a practical overlap with condition monitoring. The current transformer that measures kWh on a compressor motor can, with the right meter, also report current imbalance and harmonic distortion, which are early signs of motor and drive faults. Vibration sensors on the same motor need a higher data rate than LoRaWAN offers. Once a plant is running two sensor networks for the same assets, the second network is usually the more expensive decision, and planning one network for both from the start avoids it.

We would be candid about where private 5G does not pay. A single-site factory with ten meters, no wider digitisation plans and a cooperative electrician should install wired or LoRaWAN meters and move on. Private 5G is the right call when energy metering is one of several shop-floor workloads, or across a multi-building campus where cable runs between buildings are the main cost.

What should a factory do before the 31 December 2026 qualification date?

A factory should decide on its reference period and get its priority sub-meters live before that period starts, which for most sites means installing them this winter. ESOS requires a 12-month reference period that includes 31 December 2026 and ends before 5 December 2027. A site that wants a full year of metered data in its Phase 4 audit therefore needs the meters running by early December 2026 at the latest. The lead assessor also needs time to sign off before the 5 December 2027 deadline, so a reference period running to the end of November 2027 leaves very little slack. Starting earlier, with a period of January to December 2026 supported by whatever data exists, is the safer choice for sites that already have some metering.

For sites starting from the main meter alone, the sequence we recommend is:

1. Confirm qualification now against the 250-employee and £44m/£38m tests, across the whole corporate group, since ESOS applies at group level.
2. Pick the five or six significant loads from the most recent bills and a walk-round, before choosing any metering technology.
3. Survey the radio environment in the switch rooms and plant rooms where the meters will sit, since that decides between wired, LoRaWAN, public cellular and private network options.
4. Install and commission meters with timestamps aligned to the half-hourly supply data, so sub-meter totals can be reconciled against the bill each month.
5. Agree with the lead assessor which energy intensity ratios the meters will feed, so the data collected matches what the audit and the 2029 to 2031 progress updates will need.

Penalties for failing to carry out an ESOS assessment run to £50,000 plus £500 a day, and failure to notify to £5,000 plus £500 a day, each for up to 80 working days, with the breach published. For most manufacturers the bigger cost is a published Phase 4 action plan with savings that cannot be shown in measured kWh. We are happy to survey a site's plant rooms and switch rooms and say which metering route fits; if the answer is LoRaWAN or a cable run, we will say so.