Sanity-Checking an Installer’s Yield Claim in Ten Minutes

Every solar quote contains one number that does most of the persuading: the estimated annual generation in kWh. Everything downstream (payback, battery size, the “you’ll save £1,500 a year” line) is built on it. And it is the easiest number in the whole document to inflate, because almost nobody divides it by anything.

So divide it. The fastest solar panel quote check UK homeowners can run needs three numbers off the quote, one free web tool, and about ten minutes. It won’t tell you whether the installer is any good. It will tell you whether their generation figure is arithmetic or advertising.

Pull three numbers, ignore the rest

Open the PDF and find:

1. Array size in kWp. Usually panel count × panel wattage ÷ 1000. If the quote says “16 panels” and nothing else, that’s your first red flag and your first question.
2. Estimated annual generation in kWh. If there are two figures (an MCS estimate and a “modelled” or “expected” estimate), write down both. That pair is gold.
3. Orientation and pitch. Azimuth in degrees or a compass word, and roof pitch in degrees. Plus your postcode and any note about shading.

That’s it. Skip the panel datasheet, skip the inverter model, skip the 25-year cashflow table for now. Those matter later, at the point where you’re comparing two credible quotes on price. Right now you’re doing triage.

If the quote is a 14-page PDF and the numbers are scattered, paste it into Claude or ChatGPT and ask it to extract exactly those fields into a table. That is a good use of an LLM: reading and tabulating. Do not ask it for the irradiation value for your postcode. It will produce a confident number that is wrong by 10%, and the whole point of this exercise is that the 10% is where the lie lives.

Step one: specific yield

Divide annual kWh by kWp. You now have specific yield in kWh/kWp/year, which is the only way to compare a 4 kWp quote with a 7 kWp quote, or your quote with your neighbour’s.

Here is what the UK actually delivers, for a south-facing array at roughly 30 to 35 degrees with no meaningful shading, modelled on modern satellite irradiance data with conventional system losses:

Now derate for orientation. Relative to due south at 30 to 35 degrees: south-east or south-west costs you about 4%, due east or due west about 20%, a flat roof at 10 degrees about 8%, and due north takes 35% or more off. Modest shading (a chimney, a neighbour’s conifer, a dormer catching the afternoon) knocks off a further 10% to 20% on top of that, and it is very rarely modelled honestly in a sales document.

If a quote for a west-facing Leeds roof claims 1,050 kWh/kWp, you already know. West-facing Yorkshire has a ceiling somewhere near 740.

Step two: run PVGIS and get the in-plane irradiation

Go to the European Commission’s PVGIS PV performance tool (search “PVGIS photovoltaic geographical information system”; it’s free, no account). Drop the pin on your actual roof, tick grid-connected, set the database to PVGIS-SARAH3, choose crystalline silicon, enter your kWp, set slope and azimuth to match the quote, and leave system loss at the default 14%.

PVGIS gives you two numbers that matter. Annual energy output, and yearly in-plane irradiation in kWh/m². Write both down.

PVGIS-SARAH3   lat 51.45  lon -2.59   crystalline silicon
Installed peak power ............ 6.09 kWp
Slope ........................... 30 deg (fixed)
Azimuth ......................... 0 deg (due south)
System loss ..................... 14 %

Yearly PV energy production ..... 5 856 kWh
Yearly in-plane irradiation ..... 1 205 kWh/m2
Year-to-year variability ........ 132 kWh (2.3 %)

Changes in output due to:
  Angle of incidence ........... -2.8 %
  Spectral effects .............. +0.6 %
  Temperature and low irradiance  -5.1 %
  Total loss .................... -20.2 %

Step three: the implied performance ratio

This is the check that separates optimistic from impossible, and it takes one line in a spreadsheet:

implied PR = quoted annual kWh / (kWp × in-plane irradiation kWh/m²)

In Google Sheets, with the quote’s kWh in B1, kWp in B2 and the PVGIS irradiation in B3: =B1/(B2*B3).

Performance ratio is the fraction of the theoretical output that survives real-world losses: cell temperature, inverter conversion, cabling, soiling, mismatch, module degradation, angle-of-incidence reflection. For a new, well-ventilated, unshaded domestic UK rooftop with a decent string inverter, 0.80 to 0.86 is normal, 0.88 is about as good as it gets, and anything above 0.90 needs a written explanation. Above 1.00 the number is not a forecast, it’s an error or an invention.

Roof-integrated or in-roof panels run hotter and lose a few points. Any string with a shaded module and no optimisers loses a lot more than the shaded area suggests.

Worked example: the one that fails

A quote for a semi in Reading, RG2. Eleven 440 W panels, so 4.84 kWp. Roof faces south-west at 35 degrees. The survey notes mention the chimney shades two panels from mid-afternoon. Headline figure: “Estimated annual generation: 5,400 kWh.”

Specific yield: 5,400 ÷ 4.84 = 1,116 kWh/kWp. Already above what a perfect south-facing south-coast roof manages, on a south-west Berkshire roof with a chimney on it.

PVGIS, same plane (azimuth 45, slope 35): in-plane irradiation 1,145 kWh/m², output 4,423 kWh, which is 914 kWh/kWp. The quote is 22% higher than an unshaded model of the same roof.

Implied PR: 5,400 ÷ (4.84 × 1,145) = 0.974. That is not achievable by any rooftop PV system anywhere, let alone a partially shaded one.

And the MCS comparison is worse. MCS’s calculation (MIS 3002, following the SAP method) is simply 0.8 × kWp × S × Z, where S is the annual irradiation for your postcode zone, orientation and pitch from the standard table, and Z is the overshading factor: 1.0 for negligible shading, 0.8 for modest, 0.65 for significant, 0.5 for heavy. For this roof, S is about 1,145 and modest shading means Z = 0.8:

0.8 × 4.84 × 1145 × 0.8 = 3 548 kWh

The contractual MCS figure is around 3,550 kWh. The sales figure is 5,400. That is a 52% overstatement, and at 28p/kWh it invents roughly £520 a year of value that the roof cannot produce, which is the difference between a seven-year payback and an eleven-year one. Working out which of those it actually is belongs in a proper cost model, and the cost models and AI quote analysis pillar covers how to build one that doesn’t flatter itself.

Notice something useful in those two calculations. Unshaded, the MCS formula gives 0.8 × 4.84 × 1,145 = 4,434 kWh, and PVGIS gives 4,423. The SAP 0.8 factor and PVGIS’s default loss chain land within a quarter of a percent of each other. When two independent methods agree that closely and a quote sits 20% above both, the quote is the outlier.

Worked example: the one that passes

A Bristol quote, BS7. Fourteen 435 W panels, 6.09 kWp, due south, 30-degree pitch, no shading noted and none visible on the satellite view. The quote states two figures: MCS estimate 5,320 kWh and “our modelled estimate: 6,200 kWh.”

Reconstruct the MCS number first. The standard table value for due south at 30 degrees in the Severn zone is around 1,090 kWh/m², so 0.8 × 6.09 × 1,090 × 1.0 = 5,310. Within 0.2% of what they printed. Their MCS figure is real, calculated, and not fudged with a flattering shade factor.

Check the marketing number: 6,200 ÷ 6.09 = 1,018 kWh/kWp, and implied PR = 6,200 ÷ (6.09 × 1,205) = 0.845. Optimistic, above PVGIS’s default output of 5,856, and entirely defensible. PVGIS’s 14% blanket system loss is conservative for a brand-new install with a 97.5%-efficient inverter and short DC runs; 10% to 11% is a reasonable argument, and 10.5% is exactly what a PR of 0.845 implies.

Why the two figures differ by 17% is worth understanding rather than treating as dishonesty. MCS inherits SAP’s irradiation dataset, which is older and measurably more conservative than PVGIS-SARAH3 satellite data. An installer quoting both is showing you the floor and the expectation. An installer quoting only the high number, with no MCS figure anywhere in the document, is showing you something else.

The verdict bands, and the email

Specific yield for a well-oriented unshaded roof, adjusted for your region using the table above:

• At or below the MCS-style figure (roughly 0.9 × the regional band): conservative. You will probably beat it. Fine.
• Within the regional band, PR 0.80 to 0.86: defensible. This is what a model actually produces.
• Up to 6% above the band, PR 0.86 to 0.88: optimistic but arguable. Ask what system loss they assumed.
• PR 0.88 to 0.92: ask for the model output file before you sign anything.
• PR above 0.92, or specific yield above 1,100 outside the far south: the number is wrong. Treat the rest of the financials as equally unexamined.

Three shortcuts to failure, worth scanning for on their own: a quote that mentions no MCS estimate at all, a quote where the shading noted in the survey photos doesn’t appear as a factor in the calculation, and a 25-year savings table that multiplies year-one generation by 25 with no degradation (0.5% a year compounds to about 6% lost over the term, and the panel warranty itself tells you so).

When a quote lands in the top two bands, send one email with three questions. What azimuth, pitch and shade factor did the calculation use? What is the MCS MIS 3002 performance estimate figure? Which tool produced the modelled number (Easy-PV, OpenSolar, PVsyst) and what system loss percentage was entered?

A good installer answers all three in a paragraph, often with the Easy-PV PDF attached, because they have the file open anyway. A quote that came out of a sales script produces a phone call about the discount expiring on Friday.