Building the Business Case for Demand Reduction

There is a pattern to how demand reduction projects fail, and it is rarely technical.

The load analysis is sound. The measure is appropriate to the load shape. The kilowatts are real. The project is then written up as a simple payback period against a savings figure calculated from the wrong rate, submitted into a capital process that evaluates everything else on internal rate of return, and deferred to next year — where it is deferred again.

The engineering was never the problem.

Start with the right avoided cost

Almost every under-stated business case in this field makes the same error: it values an avoided kilowatt at the base demand rate printed in the tariff.

That is not what a kilowatt is worth. Demand-based riders stack on top of the base rate, and percentage-based taxes stack on top of those. A site that omits them can understate the benefit by a third.

Where a ratchet is in force the figure rises further, because avoiding a peak avoids the floor it would have set as well as the month it occurred in. The full derivation, including the ratchet multiplier and the seasonality, is in what a kilowatt of avoided peak is actually worth. It is worth doing once, properly, because every subsequent project at the site uses the same number.

Present it the way capital is actually allocated

Payback period is a screening tool. It ignores everything after the payback date, ignores the time value of money, and cannot rank two projects of different lives.

Capital committees know this, which is why they mostly do not use it. Presenting a payback figure into a process that runs on internal rate of return or net present value forces somebody else to do the conversion, and a proposal that has to be reworked before it can be compared is a proposal that gets set aside.

Payback is where the curve crosses zero. Everything to the right of that point is the part payback discards, and on a measure with a fifteen-year life that is most of the value.

The lines that belong in the model

Benefits. Avoided demand charges at the all-in rate. Avoided demand-based riders, if not already included. Avoided energy, where the measure reduces consumption as well as peak. Avoided or deferred capital, where freeing capacity defers a service upgrade. Program revenue, where the capability can be sold: demand response programs. Incentive payments: utility incentive programs.

Costs. Installed capital, including engineering, protection and commissioning. Ongoing maintenance. Any increase in energy consumption — storage measures deliver a service using slightly more kilowatt-hours, and that cost is real. Replacement at end of life, if it falls inside the appraisal period. The cost of measurement and verification.

Adjustments. Tax treatment, which can change the after-tax position materially: the investment tax credit and depreciation on energy equipment. Degradation, for anything whose performance falls with age.

The commonest omissions are the energy penalty on storage measures and the cost of proving the saving. Both are small; both are the kind of thing a reviewer finds and then wonders what else is missing.

There is a category of benefit that is almost always left out entirely, and it is frequently the largest one after the demand charge itself: avoided or deferred capital elsewhere. A measure that frees transformer or switchgear headroom can postpone a service upgrade that was already in the capital plan. That deferral has a present value, it is calculable from the plan you already have, and it belongs in the case with the deferral period stated explicitly rather than asserted vaguely.

The mirror image of that is worth checking too. Some measures consume capacity rather than freeing it — a battery needs a place to recharge, and the recharge is a load. If the recharge window overlaps with any period the tariff prices, or if it pushes the site toward a constraint elsewhere, that is a cost, and it is one that appears in the operating data long before anyone thinks to model it.

Say what the alternative is

A proposal that presents one option and asks for approval is asking the reviewer to do the comparison. A proposal that presents the option alongside the two alternatives that were considered and rejected, with the reason, is much easier to approve.

The alternatives are almost always the same three: do nothing, do the cheaper measure, or change the tariff.

Do nothing is not a null option and should be costed. It has a running cost — the demand charges that continue — and frequently a trend, because a site whose load is growing is a site whose peak is growing. Quantifying the do-nothing case turns a request for money into a comparison between two expenditures.

The cheaper measure should be costed even where it is inadequate, because it establishes the incremental value of the expensive one. If sequencing removes 120 kW for nine thousand dollars and storage removes 300 kW for four hundred thousand, the proposal is really asking for 180 kW at $391,000, and it should be presented that way. Reviewers who work this out for themselves tend to conclude the analysis was not done.

Changing the tariff is the option that is almost never on the list and sometimes wins. It requires no capital and no engineering, and where the site is genuinely on the wrong schedule, it can deliver a larger saving than the equipment being proposed: how to choose a rate schedule.

Choose the appraisal period deliberately

The period over which the case is assessed does more work than most people notice, and choosing it badly is a quiet way to lose a good project.

Assessed over five years, a measure with a fifteen-year life shows only a third of its benefit against its whole capital cost. Assessed over twenty-five, a measure whose control system will need replacing at year twelve looks better than it is. Neither is dishonest; both produce a misleading ranking against other proposals assessed differently.

The defensible choice is the expected useful life of the asset, with any mid-life replacement included as a cost in the year it falls, and any residual value treated conservatively. Where the organization has a standard appraisal period, use it and show the asset life alongside, so the reviewer can see what is being cut off.

There is a related trap with measures of different lives. A control change with a ten-year life and a battery with a fifteen-year life cannot be ranked on net present value alone without adjusting for the difference, and the standard corrections for that are well established. It is worth asking whoever owns the capital process which convention they use rather than picking one.

Be explicit about what could go wrong

A business case that presents a single number invites the reader to look for the flaw. One that names its own weak points is far harder to dismiss, and it is also more likely to be right.

The sensitivities worth testing on almost any demand project:

• The tariff changes. Rates are reset in rate cases. A project justified entirely on a demand rate that could fall by 15 percent should say what happens if it does.
• The load changes. A production increase, a closure, a shift pattern change. Test the case against a materially different profile.
• The measure underperforms. Storage available 92 percent of the time rather than 99 percent. Under a ratchet the difference is not proportional, because one uncovered excursion sets a floor: ratchet clauses.
• The schedule slips. A measure commissioned after the cooling season earns nothing for a year.

Sequencing, and why it changes the answer

There is an ordering question that a single-project appraisal hides.

Cheap measures should be exhausted before expensive ones, and not merely for the obvious reason. Sequencing a startup costs almost nothing and removes a peak — and if that peak is removed before a battery is sized, the battery is smaller and cheaper. A site that sizes storage against an unsequenced load buys capacity to solve a free problem and then pays for it for fifteen years.

So the appraisal of a large measure should be done against the load profile that will exist after the cheap measures are in place, not against today's. The ladder is set out in how to reduce peak demand charges.

The same logic applies to the tariff. A demand reduction that moves the site across an eligibility threshold or across the crossover load factor between two schedules produces a second benefit that belongs in the same case: how to choose a rate schedule.

Plan the proof before you spend

The single most valuable thing a first project can produce is not the saving. It is a saving that finance accepts, because that is what funds the second one.

That requires deciding, before the measure is installed, how the saving will be demonstrated: what the baseline is, what adjustments will be made for changes in production or weather, and who agrees the method. Retrofitting a verification approach after the fact produces an argument rather than a number.

The protocol conventions and the practicalities are in measurement and verification that finance will accept, and the baseline itself comes from data you should already be keeping: how to get your interval data.

The one-page version

If the case has to fit on a page, these are the lines that matter:

1. What the peak currently costs, at the all-in rate, with the ratchet effect if applicable.
2. What the measure does to the load, demonstrated on the interval series rather than asserted.
3. Capital, operating cost, and any energy penalty.
4. Net present value and internal rate of return over the asset life, with the discount rate stated.
5. The two or three sensitivities that could break it, quantified.
6. How the saving will be measured, and by when.

Six lines, all of them checkable. A proposal like that gets argued about on its merits, which is the most any project can ask for — and it survives contact with the first year's bills, which is what determines whether anyone approves the next one.