Why Your Solar 'Savings' Calculator Was Wrong From Day One
The savings number on your installer's proposal — the one that probably sold you on the system — was almost certainly wrong. Not maliciously wrong. Just optimistically wrong in three predictable ways. Three years in, the gap between that number and your actual electricity bill is real money. This is what to do about it.
What the calculator actually computed
A typical residential solar proposal contains a line that reads something like "Estimated 25-year savings: $43,200" or "You'll save $1,728 in year one." That number is the output of a calculator that took roughly these inputs:
- Your roof's square footage, tilt, and azimuth (often eyeballed from satellite imagery).
- A modeled production estimate, usually from PVWatts or a vendor equivalent.
- Your current utility rate (averaged from a recent bill).
- An assumed annual rate increase (typically 3–5%).
- An assumed annual degradation (often 0.5%, sometimes silently 0%).
- Your federal Investment Tax Credit, plus any state or utility incentives applicable at the time of signing.
The output is a multi-decade projection. The math itself isn't wrong — it's just sensitive to assumptions, and the assumptions chosen are almost always the ones that produce the larger savings number. That's not unique to solar. Every industry that sells multi-decade financial commitments — annuities, mortgages, life insurance — does this. But solar is the one where the system is physically on your roof and the gap is auditable.
The three places the calculator went wrong
1. Weather wasn't modeled honestly
PVWatts can be run two ways. The honest way uses your specific location's TMY (Typical Meteorological Year) data — roughly 30 years of measured solar irradiance and cloud cover at the nearest weather station, smoothed into an "average year" you can expect over time. The other way uses a clear-sky model — what your system would produce if every day were cloudless.
If your installer's proposal showed numbers that match what clear-sky modeling predicts, it overstated your real production by 5–15% depending on your climate. Massachusetts and New York averaged roughly 10% below clear-sky in the 2024–2025 weather window. The Pacific Northwest can be 20% below. The Southwest is closer to clear-sky.
Most reputable installers do use TMY data. But TMY is itself a 30-year average — and the last decade's weather has not been the 30-year average. The 2023 wildfire smoke season cut California production by 7–11% in affected regions. The 2024 polar vortex pattern increased Northeast winter cloud cover. These are not "edge cases" — they are happening every year now.
2. Degradation was understated
Most calculators assume 0.5% per year of capacity loss. That's the industry-standard linear degradation rate — it's roughly correct on average. The problem is the word "linear." Real degradation isn't.
Most silicon panels lose 2–3% in their first year (the so-called "light-induced degradation" burn-in), then 0.5% per year for years 2–25, then accelerate again in the last decade as encapsulant materials break down. If the calculator showed a flat 0.5% line, it implicitly assumed your panels are immortal in year 1 and youthful in year 25. Real-world cumulative loss at year 10 is closer to 7–10%, not 5%.
The honest version of this number lives in your panel's warranty document — the manufacturer guarantees something like "at least 80% of original capacity after 25 years." Work backward and that's about 0.9% per year, not 0.5%.
3. Tariff escalation was a guess that didn't age well
The calculator multiplied your year-one savings by an assumed annual rate increase to project lifetime savings. Common assumption: 3% per year. The actual U.S. residential retail electricity price grew at 4.5% in 2023 and roughly 5.5% in 2024 (EIA EPM Table 5.6.A). For most of the country this means the savings estimate is now actually LOW, not high.
Which sounds like good news. It usually isn't, because the higher rate increase is driven by the same grid pressures that are pushing utilities to restructure net metering. California's NEM 3.0 cut export credits by roughly 75% for new systems. Massachusetts's SMART program ramped down. New Jersey transitioned from SREC II to SuSI with materially different economics. Each of these changes invalidated a slice of the original calculator's projection.
What to do with the numbers you actually have
Three years in, you don't have to estimate any of this anymore. You have data. Here's the order of operations to figure out where you actually stand:
- Pull your actual production data from Enphase Enlighten or your SolarEdge portal. Sum the kWh for each calendar year since install.
- Compare against modeled expected production for the same period — pvlib's PVWatts with your system's actual specs, your latitude, your tilt and azimuth, and (importantly) measured weather for the year, not clear-sky. This is what OwlWatt does.
- Compute the cumulative shortfall in kWh.
- Multiply by your actual blended utility rate over the same period (your bills will show this; the EIA number is a fallback).
- Compare the resulting number against the savings number on your original proposal. If the gap is more than 10–15%, you probably have a documentable underperformance.
What the calculator wasn't designed to tell you
The savings calculator was a sales document. Its job was to make the financing math close. It did that job. It was not designed to be the basis for a warranty claim three years later, and it is not the right document to bring to a dispute with your installer or to an attorney.
The document you actually need is a production audit: measured production, modeled expected production with the same parameters your installer used, cumulative shortfall in kWh, dollar value at your tariff rate, and the contract clause that defines the guarantee. That's roughly what we build. It's also something you can build yourself in a spreadsheet if you have a free weekend, the patience to read your contract carefully, and a working install of pvlib.
Either way: the number on the original proposal is not the number that decides whether the system was a good deal. The data you've collected since is.