Your solar is producing far more than expected — or far less

This covers two opposite surprises that share one cause. Either a check shows a system producing significantly more than its location should allow — or the reverse, so little it would mean an essentially non-functioning system. In both cases the likelier explanation sits somewhere instinct doesn't expect: in the input, not the system.

(Looking for "less, but not extreme"? That case — a real, moderate underperformance — is covered in your solar is producing less than expected.)

Why both extremes are suspicious the same way

Expected output is computed from a model (PVGIS) with its own accuracy of roughly ±5–10 %, plus year-to-year weather variation of about ±4–6 %. Combined, that's a standard deviation around 7 % — and even with margin, it only explains a gap up to roughly 15 %.

What neither weather nor the model can explain is a gap measured in tens of per cent, in either direction. A physically healthy system cannot sustain output tens of per cent above what the sun over that particular roof allows — and equally cannot drop to a fraction of expectation while still functioning. When numbers like that show up anyway, the cause is usually a typo, not physics.

When output looks suspiciously high

In our diagnostics a rule watches for this above 115 % of expected specific yield. It doesn't report a discovered miracle — it reports an instruction to check the input.

The most common causes, in the order we see them:

• A mis-entered installed power. Typing 10 instead of the actual 8 kWp immediately inflates per-kWp output by a quarter. It's the easiest typo of all, because that single figure is typed by hand.
• A swapped unit. MWh instead of kWh is a thousand-fold gap and shows immediately; a dropped or extra zero (10× more or less) is easier to miss, because the resulting number still looks plausible at a glance.
• Wrong orientation or tilt. Entering a south orientation for a roof that's actually less favourable (or the reverse) shifts expectation away from reality.

A real explanation exists too, just rarer and never enough on its own to explain tens of per cent: slightly better-than-average weather, a small positive power tolerance versus the panel's nameplate rating, or bifacial panels gaining from ground reflection, which a general model doesn't account for. We can't tell those apart from an ordinary input mistake, though — so the finding is reported as something to check, not confirmed overproduction.

The same logic applies to the lifetime estimate from the inverter's cumulative counter: above 120 % of long-term expectation we report the same thing — check the entered power or install date. More on this kind of estimate and why we trust it less than a monthly one is in a lifetime production estimate from a single reading.

When output looks implausibly low

This is the opposite end of the same scale, and it's easier to miss — a low number surprises nobody, so it's harder to recognise it as an input problem rather than reality.

Below 20 % of annual expectation is where our diagnostics stops reporting "underperformance" and starts reporting "check the entered figure". The reason is simple: a genuinely working system, however underperforming, practically never falls below a fifth of expectation — even a badly shaded or damaged array, as long as it produces anything at all, usually stays higher. A number this low much more likely means:

• An incomplete year entered as a full one. Ten months of operation compared against a twelve-month expectation looks like a collapse that isn't real.
• A typo of an order of magnitude. Typing 900 kWh instead of 9,000 is one missing zero.
• A swapped unit — the same mistake as overproduction, just the other way.

The distinction from a genuine outage matters: an outage means a specific month produced practically nothing where normal output was expected — that's a real technical finding. An implausibly low annual total is a question about the input, not about one month.

Why we don't report it as a fault

Both of these findings carry lower severity in our diagnostics than real performance problems, and deliberately so. Asserting "overproduction" or "catastrophic underperformance" as a firm conclusion would claim more than a single extreme ratio supports — one number can't tell an input error apart from reality. So the finding always points at checking the input, not at an immediate alarm.

What to check before trusting the result

In this order:

1. Installed power (kWp). Does it match the panel's nameplate or the installation contract?
2. Units. kWh, not MWh; no extra or missing zero.
3. The period. Is the entered year actually twelve complete months, not fewer?
4. Orientation and tilt, if you entered them manually.

Run the check again after fixing the input. If the number stays extreme even after checking the input, only then is it worth considering something unusual about the system itself — but that's a minority of cases.

Where Slunora helps

The quick check reports this kind of finding separately from the real performance rules — you'll see straight away whether it's a "verify the input" prompt or a finding about the system's health. If you've connected an inverter, the same logic runs on the lifetime estimate from the very first reading.