Industrial energy sub-metering has become an easy sell. The hardware is inexpensive, the dashboards are attractive, and the pitch is reliable: you cannot manage what you do not measure. So plants buy meters, fit them to every panel with room for a current transformer, and eighteen months later the dashboard still shows data nobody acts on while the energy spend sits where it began.
The problem is not the meters. The problem is that most sub-metering projects are specified as a metering exercise when they should be specified as an accountability exercise. That distinction determines whether the project pays back in twelve months or never.
Rule 1 — Meter the cost centre, not the panel.
A panel already exists, the current transformers already fit, the cabling is short. So that is what gets metered. The plant ends up with thirty meters reporting against thirty switchboards and not a single number that maps to a person's performance.
The useful approach starts from the organisation chart, not the single-line diagram. Who carries cost accountability — the production manager for Line 2, the utilities engineer for the boiler house, the warehouse manager for the cold rooms? Each needs exactly one number from the metering system that ties to their review. If the architecture cannot deliver that one number, the architecture is wrong, however many meters are installed.
At a personal-care products manufacturer in [LOCATION], a reviewed deployment had [NUMBER] meters across the low-voltage system, none mapping cleanly to a production line because the electrical zoning had been designed for short cable runs, not line accountability. The remedy was not more meters — it was a virtual metering layer that re-aggregated existing data against production lines. Cost: roughly [AMOUNT] of engineering time. Result: a production manager who could, for the first time, be asked why one line consumed [PERCENTAGE] more energy per tonne than another making the same product.
Rule 2 — kWh per unit of output, or it does not count.
A meter reporting kilowatt-hours reports consumption. A meter reporting kWh per case, per tonne, per litre, per pallet picked reports performance. Only the second changes behaviour.
This requires the metering system to be joined, at the data layer, to production data — the manufacturing execution system, the supervisory control data, the shift log, even a manually entered case count. Without that link, the dashboard says Tuesday was a high-consumption day. With it, the dashboard says Tuesday was high because a particular line ran a particular product whose energy-per-case was [VALUE] above standard, with three candidate causes.
Plants that establish this link within the first six months consistently see [RANGE] reductions in specific energy consumption inside a year, with no capital beyond the metering itself. Plants that never join the two data streams see no measurable improvement at all. This is the single largest determinant of metering return on investment we encounter in the field.
Rule 3 — One person owns the number. Every week. Indefinitely.
This rule has nothing to do with engineering and everything to do with payback. The deployment dies the day nobody is required to explain last week's numbers in a meeting — not the consultant, not the vendor's dashboard, but a named individual inside the plant, every week, in front of their peers.
Where this is institutionalised — usually as a fifteen-minute item in an existing operations review — the metering returns its capital in [RANGE] months and continues compounding. Where the conversation does not happen, the dashboard becomes a screensaver in the engineering office and the capital is dead money. We have seen it work on a printed sheet at a 7 a.m. stand-up and fail on a costly cloud platform with mobile alerts. The institutional ritual is the active ingredient.
The honest pre-mortem
Before authorising a metering project:
- Which named individual is accountable for which metered number?
- In which existing weekly meeting is the data reviewed, by whom, against what target?
- Which production data stream is joined to the energy data so kWh-per-unit can be computed, and who owns that integration?
- What is the action protocol when a metered indicator drifts out of tolerance — who decides, who funds, who closes the loop?
If three of those four lack clean answers before the purchase order issues, the project will not pay back.
What good looks like. The best deployment we have seen in a Ghanaian plant had [NUMBER] meters, not [NUMBER]. Each served one cost centre. Each cost centre had one accountable manager. Each manager defended their energy-per-unit number weekly. Total capital was under [AMOUNT]; documented year-one savings were roughly [MULTIPLE] that figure. Fewer meters, sharper accountability, real payback.