P50 vs P90 Solar Energy Estimates: Why Two Installers Quote Different Numbers

P50 and P90 are statistical confidence levels used to describe how likely a solar generation estimate actually is to be met. A P50 estimate is expected to be met or exceeded 50% of the time, essentially the median expected outcome. A P90 estimate is deliberately more conservative, expected to be met or exceeded 90% of the time, which makes it a lower number for the same system.

Neither figure is more correct than the other, they answer different questions: P50 answers what's the most statistically likely outcome, P90 answers what's a number I can be almost certain will actually be reached, and a bank financing a large solar project cares far more about the second question than the first. A homeowner comparing residential quotes sits somewhere between those two use cases, closer to the marketing-facing P50 world in how numbers get presented, but with a genuine interest in the conservative, financing-grade question underneath it.

Why do two installers quote different generation numbers for the same roof?

If one installer presents a P50 figure and another presents a P90 figure for what is otherwise a similar system, the P50 number will look higher, purely because of the statistical basis, not because that installer's system genuinely performs better. This is one of the quieter ways quotes can be made to look more attractive without the underlying system actually being different. It isn't necessarily deliberate manipulation either, an installer used to marketing conversations might default to P50 out of habit, while one more accustomed to financing-driven discussions defaults to P90, without either one intending to mislead a homeowner comparing the two side by side.

Where does this uncertainty actually come from?

Solar generation depends on real-world weather variation year to year, which Singapore's 4.33 peak sun hours already represents as a long-term annual average, not a fixed guarantee for any single year. Some years run slightly above that average, some slightly below.

P50 and P90 estimates are two different ways of expressing how confident you can be that a specific number will actually be reached, given that natural year-to-year variation. Neither is trying to predict any one specific future year precisely, both are honest statistical descriptions of a range of plausible outcomes around a long-term average, which is genuinely the most any solar generation estimate can honestly promise a homeowner.

What does the actual P50-to-P90 gap look like in real numbers?

Solar resource assessment uses a standard statistical method: P90 sits 1.28 standard deviations below the P50 (median) value, and year-to-year irradiation variability typically runs in the region of 3-4% RMS at most locations. Combining those, a P90 figure commonly lands roughly 4-5% below the equivalent P50 figure, not a dramatic swing, but a real and precisely quantifiable one worth understanding.

Applied to a 10kWp system generating around 11,060 kWh a year at Singapore's typical 1,106 kWh/kWp yield (a P50-style figure), a P90 estimate for that exact same system might land closer to 10,500-10,600 kWh a year. That few-hundred-kWh gap is exactly the kind of small but real difference that makes two quotes for an otherwise identical system look meaningfully different on paper, purely from which specific confidence basis each installer chose to quote from.

Is this the same thing as the confidence level shown on Sunnify's own estimate tool?

No, and the shared word is a genuine source of confusion worth clearing up directly. P50/P90 is a statistical confidence level about future weather variability, how likely a stated generation figure is to actually be met given natural year-to-year swings. Sunnify's own estimate tool instead tracks data confidence, a separate ladder running from a starting rough estimate through roof-checked, layout-reviewed, to fully site-verified, reflecting how much has actually been confirmed and verified about your specific roof rather than a probability about future weather patterns.

Both are legitimate uses of the word confidence, but they answer different questions: one is about how much we know about your roof today, the other is about how variable next year's weather might be. Worth asking an installer to be specific about which one they mean if the word comes up in a conversation.

A useful way to hold both ideas at once: data confidence tells you how trustworthy today's number is given what's actually been verified about your roof, while statistical confidence tells you how much that same, fully verified number might still move around from one year to the next simply because weather is weather. Neither one substitutes for the other.

How should this change how you compare quotes?

Ask each installer directly which confidence level their quoted annual generation figure represents. If one quote uses P50 and another uses P90, adjust your comparison accordingly, or ask both installers for the same confidence level so you are comparing equivalent numbers.

This single question can meaningfully change how two otherwise similar-looking quotes actually stack up against each other. It's a short question with an outsized effect on the comparison, and a well-prepared installer should be able to answer it immediately without needing to check.

What does the payback-period impact actually look like in real numbers?

Take that same 10kWp system: at 11,060 kWh a year (P50) versus roughly 10,550 kWh a year (P90), the generation gap is about 510 kWh annually, worth somewhere in the region of S$130-150 a year at current tariffs once export credit and self-consumption are both properly factored into the calculation. Spread that across a payback calculation and it can shift the quoted payback period by a few months, not years, but enough to make two otherwise-identical-looking quotes come out meaningfully different if one quietly used the more optimistic basis.

This is a modest gap in isolation, but it compounds with every other assumption a quote makes, self-consumption ratio, degradation curve, tariff trajectory, so a P50-based quote stacked with optimistic assumptions elsewhere can end up meaningfully rosier than a P90-based quote using conservative assumptions throughout, even when the two systems being quoted are genuinely identical on paper.

Does this affect my payback period calculation?

It can, if the payback figure in a quote is built on an optimistic P50 number rather than a more conservative estimate. A payback period calculated on P50 generation will look shorter than one calculated on P90 generation for the identical system, simply because the underlying generation assumption differs.

Understanding which basis a quoted payback period uses helps set realistic expectations from the start, rather than being surprised by slower-than-quoted savings in a below-average year. A homeowner who knows their quoted payback figure is P50-based can reasonably expect roughly half of all years to underperform that number and half to beat it, which is a very different mental model from treating the quoted figure as a promise the installer has somehow guaranteed.

Further reading: see solar payback period in Singapore for how generation estimates feed into a payback calculation, and solar panel cost in Singapore for comparing quotes on price. Run the Sunnify solar estimate to see a generation figure for your own roof.