Every commercial solar business case begins with a number: expected annual generation in kWh. Boards treat it as fact; engineers know it is a model. The gap between those views causes more post-install disappointment than module failure.
Understanding how accurate solar generation estimates are at feasibility stage — and what improves them later — helps you commission analysis, read reports, and set expectations before capital is committed.
Feasibility-grade vs design-grade modelling
At feasibility, generation models use concept layout, benchmark or generic module performance, regional weather data (commonly PVGIS or equivalent for UK sites), and standard loss assumptions for inverter, cable, soiling, and shading where detectable from imagery.
At detailed design, specified module datasheets, exact string configuration, inverter efficiency curves, and site-measured shading refine the figure.
Feasibility answers “is this roof in the right ballpark?” Design answers “what will this configured system produce?”
Stage1Energy positions generation as feasibility-grade. Built from satellite and metered data. Compared against twelve months of metered output from operating commercial rooftop PV. Stage1Energy is a feasibility screening, not a precision measurement. Generation is modelled at P50 with a site-specific P90. CAPEX is quoted as a band, not a point — a firm installer quote replaces it. Returns are shown across that band, and the report states whether the recommendation changes across it. That is the right frame for commercial solar feasibility reports, consistent with our methodology and service disclaimer.
What P50 and P90 mean in practice
P50 is the central generation estimate — the figure most likely to appear in the headline kWh line. P90 is a more conservative, site-specific yield used to stress-test the case: roughly nine years in ten, output should exceed the P90 value if the model and weather file hold.
On a 400,000 kWh P50 estimate, a site-specific P90 might sit around 360,000–380,000 kWh depending on location and loss assumptions — before broader weather variability year to year.
For payback, translate generation bands into savings bandwidth. If 80% self-consumption at 25p/kWh dominates economics, a shift from P50 to P90 might move payback by months — or more if the project is marginal.
Robust investments tolerate the band. Fragile ones should be flagged in feasibility, not discovered after install.
Read how to read a solar feasibility report for stress-testing financials against generation uncertainty.
CAPEX bands and returns
Generation accuracy is only half the picture. Feasibility CAPEX uses published cost bands segmented by system size — not your installer’s quote. Returns are shown across that band, and the report states whether the pursue-or-park recommendation changes from one end to the other.
If the verdict holds across the band, you can proceed toward quotes and surveys. If it flips, you know exactly what to price before you commit — which is often more valuable than a false-precision single number.
Major drivers of error
Usable roof area. Overstated area is the classic optimism lever. Feasibility should document exclusions — setbacks, plant, shading — not only gross roof m².
Shading. Remote assessment misses winter sun angles and temporary obstructions. Nearby construction can invalidate remote shading assumptions.
Soiling and degradation. Urban industrial sites may soil faster than default assumptions. First-year output exceeds long-term average; models should show both.
Weather interannual variability. UK irradiance varies year to year. Single-year weather files do not capture full uncertainty.
Inverter clipping. Large DC:AC ratios clip peak output. Feasibility using naive kWp × factor shortcuts can overstate yield.
Export vs self-consumption mismatch. Generation accuracy is separate from savings accuracy, but boards conflate them. Wrong load profile hurts financial outcomes even when kWh is perfect.
Equipment change between feasibility and install. Substituted modules or inverter strategy shift performance. Tie feasibility assumptions to tender brief.
How providers should document uncertainty
Credible reports state:
- Modelling tool and weather data source
- Loss breakdown or combined loss factor
- Whether shading was 3D modelled or rule-of-thumb
- P50 and P90 generation where applicable
- CAPEX as a band with stated source
- Explicit note that output is a screened estimate, not guaranteed
If a report presents four decimal places without methodology, treat confidence as lower, not higher.
Compare transparency against our example report and what is in a solar feasibility report.
Checked against metered output
The strongest feasibility providers benchmark models against metered data from real commercial installs — adjusting loss factors and documenting variance.
Modelled generation has been compared against twelve months of metered output from operating commercial rooftop PV. Over the year, the difference was a fraction of the accuracy band quoted in the report. Ask any provider what back-testing they perform; marketing claims without comparable data deserve scepticism.
Post-install, compare metered output to feasibility in year one. Feeds portfolio learning and portfolio solar feasibility screening assumptions on sister sites.
Accuracy relative to installer quotes
Installers sometimes show higher yield because layouts maximise kWp or use aggressive loss assumptions — another reason to separate feasibility study from installer quote.
Independent feasibility with conservative, stated assumptions often looks lower than sales material. That gap is information, not error.
When to invest in finer modelling
Upgrade modelling depth when:
- The project is marginal on payback and small yield changes flip the verdict
- Shading is complex (urban, multi-level roofs)
- Planning or grid caps system size below geometric maximum — every kWh matters
- Investors require tighter confidence bands
For most warehouse and industrial rooftops with clear area and daytime load, feasibility-grade modelling is sufficient to decide whether to spend on surveys and design — see feasibility vs design.
Improving accuracy without over-spending
Before commissioning expensive bespoke modelling:
- Fix layout errors — confirm usable area on imagery with facilities staff.
- Obtain half-hourly load data — improves financial accuracy even when generation is unchanged.
- Align tender brief — require bidders to explain yield differences from feasibility baseline.
- Monitor after install — one season of meter data beats another remote revision.
Waiting list uses lighter modelling appropriate to filtering. Full site assessment reports at join the waiting list carry the depth needed for board papers — see solar investment board paper.
Setting board expectations
Tell directors explicitly: feasibility generation is a screened P50 estimate with a site-specific P90 stress case, not a performance guarantee. Pair headline kWh with payback sensitivity to generation, CAPEX band, and tariff.
Honest framing prevents the “underperforming asset” narrative when weather or load differs from assumptions — and builds trust when projects meet expectations.
Generation estimates are good enough for go/no-go at feasibility when methodology is transparent and readers understand the band. Perfection belongs after design, monitoring, and years of operation — not in the first report on the desk.
For current market context affecting tariffs and export value, see commercial solar feasibility in 2026.
Communicating uncertainty to non-technical stakeholders
Translate kWh bands into language finance recognises: “annual savings likely between £X and £Y at stated tariffs” lands better than a single headline kWh alone. Pair the band with the feasibility verdict so directors see that uncertainty does not automatically mean “no” — it means proceeding with eyes open and monitoring planned after commissioning.