When a solar installer tells you a system will generate 15,000 kWh per year, what they usually mean is the P50 estimate — the expected annual yield with 50% confidence. In other words, there is a 50% chance the system will produce more and a 50% chance it will produce less. For a homeowner evaluating whether solar makes financial sense, P50 is a reasonable starting point. For a bank deciding whether to lend you money, it is not enough.
What Are P50 and P90?
Solar energy yield estimates follow a normal distribution (bell curve) centered on the P50 value. The coefficient of variation (CV) measures the width of this distribution based on historical weather data variability.
- P50: The 50th percentile. There is a 50% probability that actual generation will exceed this value and 50% that it will fall below. This is the "most likely" outcome used for initial feasibility and expected ROI.
- P90: The 90th percentile. There is a 90% probability that actual generation will equal or exceed this value. This is the "conservative" estimate used for project financing and loan underwriting.
Banks use P90 because they need to be confident that the cash flows from energy savings will cover loan repayments even in below-average weather years. A loan structured on P50 would default in 5 out of every 10 years.
The P90 Formula
Where:
- P50 = expected annual yield (kWh)
- CV = coefficient of variation (decimal form, e.g., 0.10 for 10%)
- 1.282 = the z-score corresponding to the 90th percentile of a normal distribution
Example: A system in Munich, Germany with a P50 yield of 4,620 kWh/yr and a CV of 12% (0.12):
P90 = 4,620 × (1 - 0.154)
P90 = 4,620 × 0.846
P90 = 3,909 kWh/yr (84.6% of P50)
P50 vs P90 by Location
The CV varies by climate zone and local weather patterns. Arid regions with consistent sunshine have lower CV (less inter-annual variation). Maritime climates with variable cloud cover have higher CV.
| Location | Climate | P50 Yield (kWh/yr) | CV | P90 Yield (kWh/yr) | P90 as % of P50 |
|---|---|---|---|---|---|
| Phoenix, USA | Arid desert | 19,500 | 5% | 18,250 | 93.6% |
| Munich, Germany | Temperate continental | 4,620 | 12% | 3,909 | 84.6% |
| London, UK | Maritime temperate | 4,950 | 15% | 3,998 | 80.8% |
| Mumbai, India | Tropical monsoon | 5,250 | 10% | 4,577 | 87.2% |
| Sydney, Australia | Subtropical | 9,750 | 8% | 8,750 | 89.7% |
| Lahore, Pakistan | Semi-arid | 6,750 | 9% | 5,971 | 88.5% |
| Dubai, UAE | Arid | 14,025 | 6% | 12,946 | 92.3% |
| Toronto, Canada | Cold continental | 8,160 | 14% | 6,694 | 82.0% |
Why Banks Require P90 (Not P50)
Consider a bank financing a solar project with a $700/month loan payment. The system saves $800/month at P50 (generating enough power to offset the bill). At P50, the borrower can make payments. But in a low-yield year (P90 scenario), savings drop to $650/month. The borrower now has a $50/month shortfall.
Over a 10-year loan period, the probability of at least one "bad" weather year is very high. Banks underwrite to P90 to ensure the loan is serviceable in the 9th out of 10 worst-case years — not just the average year.
How CV Is Determined
The coefficient of variation comes from analysis of historical weather data — typically 10-30 years of satellite-derived or ground-measured solar irradiance (GHI) data.
- Low CV (5-8%): Arid and subtropical climates (Phoenix, Dubai, Perth). Consistent year-to-year solar resource.
- Medium CV (8-12%): Continental and Mediterranean climates (Madrid, Milan, Sydney, Johannesburg). Moderate variability.
- High CV (12-18%): Maritime and monsoon climates (London, Mumbai, Tokyo, Vancouver). High year-to-year cloud cover variation.
Data sources include NASA POWER, Solargis, Meteonorm, and local meteorological stations. Free tools like PVGIS and NREL PVWatts provide P50 estimates but may not report CV directly. Specialist tools like PVsyst, Helioscope, and SolarGIS Pro provide full P50/P90 analysis.
P50 for Homeowners, P90 for Financing
For a homeowner paying cash, using P50 for ROI calculations is reasonable — you are the equity investor and can absorb below-average years. However, if you are taking a solar loan, your lender will apply P90. Here is how the numbers differ for a typical system:
| Metric | Using P50 | Using P90 | Difference |
|---|---|---|---|
| Annual Yield (Germany 3.5 kWp) | 4,620 kWh | 3,909 kWh | -15.4% |
| Annual Savings (€0.32/kWh) | €1,478 | €1,251 | -15.4% |
| Payback Period (€7,000 cost) | 4.7 years | 5.6 years | +0.9 yr |
| 10-Year ROI | 111.1% | 78.7% | -32.4% |
| 25-Year Profit | €29,950 | €24,275 | -19.0% |
See our 10-year solar ROI guide for a full sensitivity analysis between P50 and P90 scenarios across different markets.
What If Your Actual Yield Differs from P90?
P90 means that in 9 out of 10 years, you will achieve at least that yield. In the 10th year, you may fall below. Mitigations include:
- Oversizing the array 5-10% — additional panels cost little but buffer against low-yield years.
- Operations and maintenance — keeping panels clean and inverters well-ventilated recovers 3-8% of potential losses.
- Tariff escalation — if grid rates rise 3-5% annually (which they have historically in most markets), the financial impact of lower yield is offset by higher savings per kWh.
How to Get P50 and P90 for Your Location
You can obtain your P50 yield estimate from free tools like PVGIS (EU/Global), NREL PVWatts (US), or APVI (Australia). These tools use historical satellite irradiance data and account for panel orientation, tilt, and system losses.
For P90, you need the CV for your location. General CV ranges by climate are given above. For a more precise calculation, use PVsyst or Helioscope which apply site-specific weather files from 15-30 year databases.
Use Our Free P50/P90 Yield Calculator
Our interactive tool calculates both P50 and P90 for any location. Enter your address, system size, and panel orientation to get bankable yield estimates ready for financing applications.
Calculate P50 and P90 for your system
Open the Yield Calculator →Data sources: NASA POWER (2025), Solargis, Meteonorm 8.1, PVGIS 5.2, NREL PVWatts v8. CV estimates are indicative. Precise P90 analysis requires site-specific weather data and should be performed by a qualified solar engineer.
Last updated: July 2026 | Browse all guides