Backup or Bill Savings: Two Battery Jobs, Two Prices

A battery sold for resilience and a battery sold for bill savings are two different products that happen to share an enclosure. They need different hardware, they are sized by different arithmetic, and only one of them produces cash.

Quotes rarely say so. Mine priced hardware whose entire purpose was to work during a blackout, then quoted an annual savings figure computed on the assumption that the grid stayed up all 8,760 hours of the year, and ran both into one payback number. Do that and the resilience half is valued at zero by default. You paid for it; the spreadsheet pretends you didn't.

Both lines have names you can look for. The blackout hardware tends to appear under a product name rather than a function, so it is the product page that tells you what you bought: Enphase describes its IQ System Controller 3 as "automatically detecting and seamlessly transitioning the home energy system from grid power to backup power in the event of a grid failure" (Enphase IQ System Controller 3, read 18 August 2026), and next to that box sit the critical-load subpanel and the electrician hours that fill it. The savings line has its own cluster of inputs: a rate schedule name, an escalation rate, a reserve percentage. Find both clusters on your own quote before you read the payback figure, which was built from the second and charged for both.

A regulator that pays different amounts for each job

You do not have to take my word that these are separate products. California's Self-Generation Incentive Program prices them separately, on one table, in public.

On the CPUC's SGIP page (read 18 August 2026), the residential storage incentive rates run like this: Residential Solar and Storage Equity — storage at $1,100/kWh, Equity Resiliency — $1,000/kWh, San Joaquin Valley Residential — $1,100/kWh, and Small Residential Storage — $150/kWh (CPUC, Self-Generation Incentive Program).

Read the spread. The resiliency-oriented budgets pay roughly seven times what the general small residential storage budget pays for the same kilowatt-hour of the same chemistry. Nothing about the cells changed. What changed is which job the ratepayer money is buying: keeping a household powered through a Public Safety Power Shutoff in a high fire-threat district is worth something to the system that ordinary evening load-shifting is not.

Two practical notes before that table enters anyone's arithmetic. Three of those four budgets carry the label available through 2025 on the CPUC page itself, so a 2026 quote that nets out SGIP money is making a claim you need to verify — the program administrators publish live budget status, an incentive step tracker and the 2026 SGIP Handbook at selfgenca.com. And if you are outside California, the point is not the numbers. The point is that a large public program looked at these two uses and declined to price them the same.

The resilience job: hardware, kilowatts, and no revenue line

A battery that only shaves your bill can sit alongside everything else and never disconnect from the utility. A battery that carries you through an outage has to do something quite different — separate your house from the grid, form its own small island, and hold voltage and frequency inside it. That separation is a physical device, and it is the line item people skim past.

Depending on the design it appears as an automatic transfer switch, a gateway, or what installers call a microgrid interconnect device. Often it also means a critical-load subpanel: a second breaker box holding only the circuits you intend to keep alive, plus an electrician's hours relocating those circuits into it.

The code framework sits mostly in the National Electrical Code. Article 706 covers energy storage systems above a small capacity threshold; Article 702 covers optional standby systems and requires the standby source to have capacity for the loads intended to run simultaneously; and 705.13 permits a listed energy management system to limit power flow as an alternative to some panel or service upgrades. Read the section text rather than a summary of it — NFPA publishes the code for free reading at nfpa.org free access. One caution matters more than the section numbers. The NEC is republished on a three-year cycle, but it is not law anywhere until a jurisdiction adopts it, and adoption runs years behind publication. Ask your permit office which edition it enforces before you quote a section number at anybody — the answer is a local fact, and the newest book on the shelf is usually the wrong one.

Two things follow for your arithmetic.

The binding constraint here is kilowatts, not kilowatt-hours. You can own 20 kWh and still be unable to start the one appliance you bought the battery for, because a compressor's starting surge is answered by the continuous power rating and not by the energy behind it. So the resilience spec is a list of circuits and a kW figure, decided before anyone talks capacity. The energy half is a separate exercise on your own interval data.

There is no revenue line. Backup does not generate money. It avoids loss, and only on the days the grid fails. So a backup battery has no payback period in the ordinary sense, because the denominator does not exist. If a proposal shows one figure in years for a backup-configured system, it computed the arbitrage job and handed you the resilience hardware as a rounding error.

The self-consumption job: arithmetic on your own tariff

This side does have a revenue line, and it is computable to the cent from documents you can download today.

Per cycle, the battery delivers d kWh into your house during an expensive window, having absorbed d ÷ RTE kWh earlier. So:

value per cycle = d × p_peak − (d ÷ RTE) × p_charge
                = d × (p_peak − p_charge ÷ RTE)

Both prices come off your rate schedule, and where to find each one is worked through in the sizing post. The case worth pulling out here is the second one: when the battery charges from your own array, p_charge is not a bill line at all. It is the export credit you declined to collect.

Which is where California's net billing tariff enters. Under Decision 22-12-056, adopted 15 December 2022 and applying to interconnection applications submitted on or after 15 April 2023, exports are compensated at hourly avoided-cost values rather than at the retail rate (CPUC, Net Billing Tariff, read 18 August 2026). When the credit for a midday export hour is small and the import price for a 7 p.m. hour is large, the gap between them is what a battery harvests. That gap is the whole economic case for self-consumption, and it is set by a tariff sheet that has nothing to do with which battery you buy.

Where the two jobs fight: the reserve setting

Here is the piece most quotes never show, and it is one line of arithmetic you can do yourself.

Call U the usable capacity and r the state-of-charge floor you hold back so the battery is never empty when the grid drops. Only U × (1 − r) is available to cycle. So the reserve has an annual price:

cost of reserve = U × r × (p_peak − p_charge ÷ RTE) × N cycles per year

Worked with placeholder figures so the shape is visible — these are not your rates and not a forecast:

• U = 13.5 kWh usable · r = 0.30 · RTE = 0.88 · N = 300 cycles
• p_peak $0.45/kWh · p_charge $0.12/kWh
• Effective spread: 0.45 − (0.12 ÷ 0.88) = 0.45 − 0.136 = $0.314/kWh
• Reserve held: 13.5 × 0.30 = 4.05 kWh
• Cost of reserve: 4.05 × 0.314 × 300 = about $381 per year

Now turn that reserve into hours. At an average backed-up load of 0.6 kW, 4.05 kWh runs roughly 6.75 hours. So a standing 30% reserve buys about seven hours of readiness at something near $56 per hour of coverage, per year.

That is a number you can actually decide against. Seven hours of guaranteed lights, fridge and internet for $381 a year is either obviously worth it or obviously not, depending on how often your feeder goes down and what an outage does to your household. Utilities model this at scale with tools like Lawrence Berkeley National Laboratory's ICE Calculator. A household does better with the concrete version of the question: what does one evening without power cost me, and how many of those do I get?

And notice the dial. Reserve is usually a slider in the manufacturer's app, adjustable after commissioning. The same battery can be mostly an arbitrage machine in June and mostly an insurance policy in storm season. Nobody mentions that at the kitchen table, because proposal software models one setting for twenty years.

Run the same line at other settings and the dial turns into a price list. Holding the placeholders above — 13.5 kWh usable, a $0.314 effective spread, 300 cycles — a 10% floor costs about $127 a year and holds roughly 2.25 hours at 0.6 kW, while a 50% floor costs about $636 and holds about 11.25 hours. The spread moves with the calendar too: time-of-use schedules commonly define separate summer and winter periods at different peak prices, so read the season definitions on your own schedule before letting one annual figure stand for both halves of the year.

One credit that used to blur the line, and no longer does

Until recently the federal Residential Clean Energy Credit softened this whole argument, because it applied to storage regardless of which job the storage did — battery storage technology of at least 3 kWh became eligible from 2023, with no requirement that it be paired with solar or that it ever save a cent on a bill. Pure insurance got the same 30% as pure arbitrage.

That is over. The IRS page states the credit "is not available for any property placed in service after December 31, 2025" (IRS, Residential Clean Energy Credit, read 18 August 2026). Resilience spending no longer gets a federal discount, which raises the bar the backup job has to clear on its own merits. State and utility programs may still exist. Federal does not, and a proposal that says otherwise is telling you something about how carefully it was assembled.

Make the quote say two numbers

Ask for the same battery priced twice: once with full backup capability, once as a grid-tied self-consumption unit with no transfer equipment and no critical-load subpanel. Most installers can produce both in an afternoon, and the difference between the totals is the price of resilience with nothing else mixed in.

Then, holding that figure, three follow-ups. Which circuits are in the backup panel, and what continuous kW rating has to start all of them? What reserve percentage did the savings estimate assume, and what does the annual number become at 40%? And which line of my rate schedule produced the p_peak in that estimate?

Add a fourth ask that costs nobody anything: the warranty document for each box, not for the system. The two halves do not age together — Enphase publishes a 15-year limited warranty up to 6,000 cycles on the IQ Battery 5P and a 10-year limited warranty on the IQ System Controller 3 (enphase.com, read 18 August 2026), so on that equipment the device that makes backup possible carries the shorter term. Whatever brand your quote names, those numbers sit on the manufacturer's own pages and the installer does not set them.

If the answers come back as a single blended payback figure, the proposal was not sized on your house or your tariff. It was sized on a template.