Every solar quote carries a performance warranty, and it is nearly always drawn as a straight line: some percentage in year one, a fixed decline every year after, a guaranteed floor at year 25. It is a clean promise and it is easy to compare between manufacturers.
It is also a commercial undertaking rather than a measurement. What panels actually do has been measured, repeatedly, and the results are public.
What the field data says
The reference work is a compendium assembled at the US National Renewable Energy Laboratory. It aggregated more than 11,000 degradation rates from almost 200 studies across 40 countries, then re-analysed them to correct for the sampling bias that creeps in when large studies and small ones are pooled naively.
For crystalline silicon, the technology on essentially every domestic roof, it reported median degradation of 0.5 to 0.6% a year, with the mean between 0.8 and 0.9%.
So there are two honest answers to what a panel loses, and which one applies to you is not knowable in advance. The median describes the panel you probably have. The mean describes the fleet, including the ones that go wrong.
What those rates compound to
Degradation is multiplicative. Losing 0.5% of a smaller number each year is not the same as subtracting a fixed amount, and over 25 years the difference is visible.
| Annual rate | After 10 years | After 15 years | After 25 years |
|---|---|---|---|
| 0.5%/year (field median, best case) | 95.1% | 92.8% | 88.2% |
| 0.6%/year (field median, worst case) | 94.2% | 91.4% | 86.0% |
| 0.85%/year (field mean) | 91.8% | 88.0% | 80.8% |
| Typical linear warranty floor | 93.5% | 91.0% | 86.0% |
The first three rows compound the stated rate. The fourth is arithmetic on a common warranty shape: 98% guaranteed at the end of year one, declining 0.5 percentage points a year thereafter. Warranty terms differ between manufacturers and yours must be read rather than assumed.
Read the bottom two rows against each other. A typical linear warranty floor at year 25 lands at 86.0%. The measured fleet mean lands at 80.8%. The warranty is not set generously below what panels do; it is set close to the median, which means the modules in the bad tail are precisely the ones that will breach it.
That is not an accusation. It is what a warranty is for. But it does mean the guaranteed floor should be read as the level at which you get to make a claim, rather than as a conservative estimate of performance.
The first year is different
Most warranties allow a larger drop in year one than in any year after it, commonly two or three per cent against half a per cent subsequently. That is real and it has a physical cause. Crystalline silicon modules undergo initial light-induced degradation in their first hours and weeks of exposure, and the module stabilises at a level below its factory flash test.
It also means a year-one performance check against the nameplate rating will look disappointing even on a perfectly good array, which is worth knowing before anyone panics.
What actually makes a panel degrade faster
The compendium found that hotter climates and sustained high temperatures were associated with higher degradation in some products. For a British roof that is a mild consideration rather than a dominant one. Two others matter more.
- Thermal cycling and humidity working on the laminate, the junction box seals and the cell interconnects. This is where most long-run failures originate, and it is mechanical rather than electrical.
- Mounting that traps heat. A roof-mounted array runs hotter than a ground-mounted one with free airflow behind it, which is already visible in the annual loss budget before any question of ageing.
Most degradation curves in the data were close to linear. The exception was the worst-performing modules, which showed non-linear wear-out, meaning they held up acceptably and then fell away. That shape is what makes a levelised cost calculation over 25 years sensitive to which tail of the distribution a product turns out to be in.
How to use this in a financial estimate
If a quote projects savings over 25 years using a flat annual generation figure, it is overstating the later years. If it applies the warranty line, it is using a commercial floor as a central estimate, which understates the median case.
The defensible approach is to model the median and state the mean as the downside, which is the same reason every figure on this site carries a lower bound alongside the expected value. What that looks like in practice, and how to check whether a quote has done it, is covered in what a 25-year warranty actually covers.
Our calculator works on current-year output rather than projecting a 25-year revenue stream, which avoids compounding an assumption about degradation on top of an assumption about tariffs. Where it does project forward, it reports the range.
Jordan, D., Kurtz, S., VanSant, K. and Newmiller, J., Compendium of photovoltaic degradation rates, Progress in Photovoltaics: Research and Applications, 24(7), 2016. More than 11,000 degradation rates from almost 200 studies in 40 countries; median for crystalline silicon 0.5–0.6%/year, mean 0.8–0.9%/year. Checked 2026. Source