In short
- Efficiency reduces the area under the load curve — fewer kilowatt-hours for the same output. It saves on the energy line.
- Shifting changes the shape without changing the area — the same kilowatt-hours, arriving at different times. It saves on the demand line and on time-of-use energy pricing.
- The two are funded differently, measured differently and paid for differently by utility programs.
- Shifting can slightly increase total consumption because storage has losses, and that cost belongs in the model.
- The best measures do both. The dangerous assumption is that any measure doing one must be doing the other.
Both are worth doing. Confusing them is what produces a savings projection that misses by half and an energy manager who has to explain why.
The geometric distinction
Put the load curve on a chart. Efficiency pulls the whole curve downward: less power drawn to do the same work, at whatever hour it happens. Shifting pushes part of the curve sideways: the same total work, redistributed in time.
A demand charge prices the height of that curve. An energy charge prices the area. Efficiency reduces the area and reduces the height only where it acts. Shifting reduces the height and leaves the area alone.
That is the whole distinction, and every practical consequence follows from it.
Why efficiency often fails to cut the demand charge
Take a lighting retrofit that halves lighting consumption at a plant whose peak occurs at two in the afternoon during a production surge.
The energy saving is real and continues every hour the lights are on. The demand saving is limited to whatever share of the 2 p.m. interval average the lights represented — frequently small, and at a daylit facility potentially close to zero if the lighting was dimmed at that moment anyway.
The retrofit is a good project. It is an energy project. Presented as a demand project it will disappoint, and the disappointment attaches to the next proposal too. More examples of the same trap are in six efficiency measures that do not cut your demand charge.
Why shifting often fails to cut the energy bill
The reverse holds. Moving a batch process from afternoon to night, on a flat energy tariff, changes the kilowatt-hours not at all. The demand charge falls; the energy charge does not move.
On a time-of-use tariff, shifting does reduce the energy cost, because the same kilowatt-hour is priced differently by period. That is a second, separate saving, and whether it exists at all depends on which schedule you are on: time-of-use rates for commercial accounts.
And where storage is involved, shifting increases consumption slightly. A thermal storage system or a battery returns less energy than it absorbs, so the same service costs a few percent more kilowatt-hours. It is a genuine cost and it belongs in the model rather than in a footnote.
Sorting your own options
| Measure | Cuts kWh | Cuts peak kW | Notes |
|---|---|---|---|
| Lighting retrofit | Yes | Only if lights run during the peak interval | Strong energy case, weak demand case |
| Variable speed drive on a fan or pump | Yes | Usually yes | One of the few that reliably does both |
| Insulation and envelope work | Yes | Partly, by reducing cooling load | Slow, durable, works on both |
| Staggered startup sequencing | No | Yes | Free; changes shape only |
| Pre-cooling | No | Yes | May slightly increase kWh |
| Thermal storage | No, slightly negative | Yes | Round-trip losses |
| Battery peak shaving | No, slightly negative | Yes | Round-trip losses |
| Compressed air leak repair | Yes | Yes, if compressors run at the peak | Cheap, and usually both |
The two rows worth dwelling on are the drives and the leak repair. Both reduce consumption continuously and reduce the load during whatever interval sets the demand charge, because the equipment involved is generally running at the peak. Measures like these are the ones to find first, because they earn on both lines and need only one business case.
Getting the projection right
The discipline is simple and rarely followed: apply the proposed change to the interval series, then recompute both determinants from the modified series.
Not a percentage applied to the monthly total. Not an assumption that a 20 percent energy saving produces a 20 percent demand saving. Interval by interval, then recompute the highest average, then price it.
That method catches every failure described above, and it takes an hour once the data exists: how to get your interval data. It also produces the baseline that any later savings claim will be measured against, which matters more than most people expect: measurement and verification.
Which one your site should do first
Read the split off your own bill. Total the energy line and the energy-based riders; total the demand line and the demand-based riders. Whichever is larger is where the leverage is.
A high load factor site running continuously has a bill dominated by energy and comparatively little shape to exploit — efficiency first. A low load factor site with sharp peaks has a bill where demand and its riders can rival or exceed energy, and the cheapest measures available are shape measures: how to reduce peak demand charges.
Most sites should do both eventually. The order is decided by the bill, not by which project sounds more impressive.
One caution on sequencing the two. Efficiency measures change the load shape as well as its size, so a shifting strategy designed around a load profile that a retrofit is about to alter may be aimed at a peak that will no longer be there. Where both are planned, model the efficiency measure first, then design the shifting strategy against the resulting profile. Doing it the other way round produces a controller tuned to a building that no longer exists.
