One table per mounting. Watch how little the order moves and how much the figures do.
Due south, 35 degrees, unshaded. The best case anybody can quote you.
Due south at 30 degrees, standing on the ground. Notice that the whole column has barely moved from the one above it.
Across all thirteen points the ground-mounted figure tracks the roof figure almost exactly. Nobody needs a roof to get roof-level output. They need south and they need a clear view of the sky.
Vertical, which is what a wall bracket gives you. Every figure drops by roughly a quarter to a third.
The arrangement a great many people end up with, because it is the wall they happen to have. It costs about half.
Top to bottom of the first table is a spread of about 43 per cent. Top of the first table to top of the last is a spread of about 112 per cent. The decision you make about mounting is several times larger than the one geography already made for you.
| Caution | Why it bites |
|---|---|
| Unshaded | Every figure assumes a clear view of the sky. A neighbouring roof, a chimney or a mature tree removes more than any of the differences shown above. |
| One point per region | A coastal address and an inland one inside the same region do not perform alike, and neither is the city we queried. |
| Output, not saving | A kilowatt hour counts only if you are using it when it arrives. Storage is outside the approved class, so nothing is held over. |
| One device | The connection rules cap a household at a single unit, so these are the totals rather than a starting point to be multiplied. |
Method, so anyone can repeat it: PVGIS API v5_2, peak power 0.8 kWp, system losses 14 per cent, crystalline silicon, building-integrated, queried 13 September 2026.