The electricity line in most operating budgets is produced by taking last year's total and adding a percentage. It is quick, it is defensible in the room, and it is wrong for a specific and predictable reason.
An industrial electricity bill is not one number. It is four components that move independently: energy volume, billed demand, rider rates and taxes. Applying a single growth rate to their sum assumes they all moved together, which they did not, and the error compounds because the components have different relationships to production.
Forecast the determinants, not the total
The alternative is not much harder and is considerably more useful.
Energy volume is the component that tracks production, and it is the one most people already forecast well. Take the production plan, apply the relationship between output and consumption from historical data, and adjust for known efficiency measures arriving during the year.
Billed demand does not track production in the same way, and this is where most budgets go wrong. A site running 15 percent more volume through the same equipment may show almost no increase in peak demand — or a large one, if the extra volume is achieved by running more equipment simultaneously rather than for longer. Forecast it from billed demand by month, adjusted for specific known changes, rather than by scaling consumption.
Rider rates are set administratively and change on their own schedules, some quarterly, some annually. They are knowable in advance more often than people assume.
Taxes follow whatever is beneath them.
Building the forecast this way takes an afternoon and it produces something else worth having: a variance analysis that identifies which component moved when the actual bill arrives, rather than an unexplained gap.
The ratchet makes demand partly pre-determined
There is a useful asymmetry here. Where a ratchet clause is in force, part of next year's billing demand is already fixed by peaks that have already happened.
If the site's summer peak sets a floor at 80 percent for the following eleven months, then the demand determinant for a substantial part of the budget year is not a forecast at all — it is a known quantity, sitting in data you already hold. That is unusually convenient for a budget line, and it also gives an exact date on which the position improves, if no new peak intervenes: ratchet clauses.
The corollary is that a peak occurring in the current year is a budget event for the next one, and should be reported as such at the time rather than discovered in the following March.
Rate cases are knowable
A general rate case is a public proceeding at the state commission, with filings, a schedule and a proposed effective date. Riders are updated by public filings too.
A budget that treats a rate change as an unforeseeable event is not using information that is freely available. Checking the commission's docket for pending cases affecting your utility, once a year at budget time, converts a variance into an assumption. If a case is pending with an uncertain outcome, that is exactly what a sensitivity range is for.
Build a range, and know what drives its width
A single number invites the question of whether it is right; a range with named drivers invites the more useful question of which driver matters.
The variables worth ranging:
- Production volume, which drives energy.
- Peak demand, which is usually the widest single uncertainty and is only loosely coupled to volume.
- Fuel and purchased power adjustment, which can move materially and is outside anyone's control.
- The outcome of any pending rate case.
- Weather, where the load has a significant cooling or heating component.
Presenting the budget as a base case with two or three quantified sensitivities is both more honest and, in practice, easier to defend than a point estimate, because the conversation moves from whether the number is right to which of the drivers the business wants to hedge.
Where projects belong in the budget
Two rules that keep the budget honest and, incidentally, keep energy projects credible.
Only include savings from measures that are actually committed — approved, funded and scheduled. A budget that assumes savings from a project still awaiting approval creates a variance that will be attributed to energy management rather than to the approval process.
Phase them realistically. A measure commissioned in September earns four months of saving in the budget year, not twelve, and a summer-only measure commissioned in September earns none at all. Under a ratchet, a measure may earn nothing until an existing floor expires — which is a real effect and one that should be in the phasing rather than discovered later.
The value to apply to those savings is the same figure every other calculation on this site uses: what a kilowatt of avoided peak is actually worth.
The monthly habit that makes it work
A budget produced once a year and compared once a year is a document. The version that is useful is checked monthly, against the four components separately.
Chart billed demand, energy, load factor and blended cost per kilowatt-hour by month. Compare each against the forecast. When something diverges, the chart tells you which component did it, which is most of the diagnosis — and a step change in any of them with no operational explanation is the earliest signal that something structural has moved, whether that is a tariff change, a control system that stopped working, or a metering problem: a 12-point audit for any commercial electricity bill.
Two minutes a month, and it turns the annual budget conversation from an explanation into a report. The underlying skill is the same one everything here depends on, which is being able to read the bill as a set of determinants rather than as a total: how to read an industrial electricity bill.
