Two numbers appear in every serious yield report, and one of them is misread almost universally.
Teams reviewing yield assumptions can keep the engineering model separate from the record of review work; this guide is one option for documenting time and handoffs.
Reviewed August 9, 2026.
What each one means
P50 is the median annual yield. In any given year the plant has a 50% probability of producing more than this, and a 50% probability of producing less. It is the central estimate and the number used for project planning.
For independent background on photovoltaic performance, resource data and current industry practice, consult Global Solar Atlas.
P90 is the level the plant has a 90% probability of meeting or exceeding. It is the conservative figure, and it is what lenders size debt against.
The misreading
P90 does not mean ninety per cent of P50.
It is a probability of exceedance, not a proportion. A plant with a P50 of 1,700 MWh and a P90 of 1,588 MWh has a P90 that happens to be 93% of P50 — but that ratio is an output of the calculation, not its definition.
Reading P90 as "90% of P50" produces two errors at once. It understates the downside when uncertainty is high, and it overstates it when uncertainty is low — and in both cases it hides the thing the pair of numbers was there to tell you.
How they relate
Through the combined uncertainty of the assessment.
At roughly ±6% total uncertainty, P90 sits about 7.7% below P50.
At roughly ±10%, P90 sits about 12.8% below.
The gap widens as uncertainty rises, which is the whole mechanism. Two plants with identical P50 figures and different P90s do not differ in expected output — they differ in how well the site and the data are known.
The ratio to look at
P90 divided by P50, typically 0.85 to 0.92 depending on location and data quality.
Above about 0.92 generally indicates good irradiance data, real site measurement and a conventional technology stack.
Well below indicates the opposite, and it is worth asking why before asking anything about the equipment.
That ratio is the single most useful number in a yield report and it is not printed — you divide two figures that are.
What sits around them
P75, P95 and P99 appear in some reports, defined the same way at their respective probabilities. P99 is a near-worst-case figure.
Performance guarantees are frequently set at about 95% of P90, which is a different number again and is what the contract actually promises.
And degradation applies on top, at 0.5 to 1% per year, so a P50 for year one is not a P50 for year fifteen.
The two questions to ask
"What combined uncertainty does this assessment assume?" It is the input that produces the gap, and a report that cannot state it has not done the calculation properly.
"Is this P50 or P90?" Sounds trivial. A savings figure built on P50 and a debt schedule built on P90 describe the same plant and imply different outcomes, and proposals mix them more often than they should.
The short version
- P50 is the median: 50% probability of being exceeded in any year. P90 is the level with a 90% probability of being met or exceeded
- P90 is not ninety per cent of P50 — it is a probability of exceedance, and the ratio is an output rather than a definition
- The two relate through combined uncertainty: ±6% puts P90 about 7.7% below P50, ±10% about 12.8% below
- The P90/P50 ratio typically runs 0.85 to 0.92; above 0.92 indicates good data and real site measurement
- Performance guarantees are frequently set near 95% of P90, which is a third number again
- Ask what combined uncertainty the assessment assumes, and whether any figure you are shown is P50 or P90