Method and sources

How the numbers are worked out

In short: we model your panels' output for every hour of a typical year at your postcode, using the EU's PVGIS sunshine data and PVGIS's own panel model. We run that against real UK homes' electricity use, every 10 minutes. What a home uses as it's generated is the saving. What it doesn't use is usually given away.

1. Sunshine at your postcode

We take PVGIS's typical meteorological year (PVGIS-SARAH3 satellite radiation and ERA5 weather) for a 0.5° grid over Great Britain: 179 cells. A typical year is stitched from whole months of real years, so its totals can sit several percent off the long-term mean. We therefore rescale each month to PVGIS's 2005–2023 average at the same point. Your postcode district then gets its own long-term sunshine: we scale its nearest cell's year by the ratio of PVGIS's long-term figure at your district to the figure at the cell. On the Cumbrian coast, for example, that is 16% more than the cell up on the fells.

PVGIS gives each hour's irradiance at the satellite's scan time, about 11 minutes past the hour, and that is when we place the sun.

2. What the panels produce

Sun position follows NOAA's solar position algorithm (after Meeus). Light on the panel is direct, sky and ground-reflected light:

G_panel = DNI·cos θ + DHI·[(1 − F1)(1 + cos β)/2 + F1·a/b + F2·sin β] + GHI·ρ·(1 − cos β)/2

Here θ is the angle between sun and panel, β the tilt, and ρ = 0.2 the ground's reflectance. F1 and F2 come from Perez et al. (1990), the "allsitescomposite1990" coefficients. Reflection losses use Martin and Ruiz (2001), with a_r = 0.16:

IAM(θ) = (1 − e^(−cos θ / 0.16)) / (1 − e^(−1/0.16))

The module model is PVGIS's own, from its "Data sources and calculation methods" page. Module temperature follows Faiman, with PVGIS's coefficients for free-standing crystalline silicon. Power follows Huld et al. (2011), with PVGIS's c-Si coefficients:

T_m = T_air + G_eff / (26.9 + 6.2·wind) P_dc = G′·P_rated·(1 + k1·ln G′ + k2·ln²G′ + k3·T′ + k4·T′·ln G′ + k5·T′·ln²G′ + k6·T′²) G′ = G_eff/1000, T′ = T_m − 25 P_ac = min(P_dc · (1 − 0.14) · c(tilt, direction), inverter limit)

0.14 is PVGIS's standard allowance for cabling, inverter, mismatch and dirt. The inverter limit is the kit's stated output, never above the legal 800 W. Every one of these formulas is tested against the pvlib library's implementation.

Calibration to PVGIS

With the steps above we match PVGIS within about 1% for south-facing and flat panels, but upright east- and west-facing panels came out 4–8% high. We checked terrain horizons (no effect) and the typical year's mix of direct and diffuse light (the same as the long-term mean). Neither explains the gap. PVGIS is the reference most people check against, so we scale each panel direction and tilt by PVGIS's own ratio, measured at 12 GB locations. For upright panels the factor is:

Facing	N	NE	E	SE	S	SW	W	NW
Upright (90°)	0.811	0.887	0.951	0.981	0.999	0.988	0.962	0.901
Tilted 35°	1.078	1.052	1.018	0.999	0.994	1.003	1.024	1.059

How close we get

We test on six towns that are not among the 12 calibration sites, with six setups each, two of them between the calibrated angles. The comparison is PVGIS's yearly output for 1 kW of panels at the postcode district, against ours. The largest difference is 4.2% and the average is 1.5%.

District	Setup	PVGIS kWh	Ours kWh	Difference
SW1A	railing S	743	743	−0.0%
SW1A	railing SW	670	677	+1.1%
SW1A	railing E	522	517	−0.8%
SW1A	garden S	1019	1024	+0.5%
SW1A	tilt 50 SSW	989	999	+1.0%
SW1A	tilt 25 ESE	895	895	−0.0%
BS6	railing S	743	724	−2.6%
BS6	railing SW	688	673	−2.3%
BS6	railing E	511	525	+2.8%
BS6	garden S	1028	1012	−1.6%
BS6	tilt 50 SSW	1004	984	−2.0%
BS6	tilt 25 ESE	892	896	+0.4%
M1	railing S	646	619	−4.2%
M1	railing SW	588	582	−1.0%
M1	railing E	458	446	−2.8%
M1	garden S	886	867	−2.2%
M1	tilt 50 SSW	861	843	−2.1%
M1	tilt 25 ESE	782	773	−1.2%
NE1	railing S	719	717	−0.3%
NE1	railing SW	649	644	−0.8%
NE1	railing E	476	481	+1.1%
NE1	garden S	950	947	−0.3%
NE1	tilt 50 SSW	933	931	−0.3%
NE1	tilt 25 ESE	813	815	+0.2%
G1	railing S	637	620	−2.8%
G1	railing SW	588	576	−2.0%
G1	railing E	435	427	−1.8%
G1	garden S	859	845	−1.6%
G1	tilt 50 SSW	842	829	−1.6%
G1	tilt 25 ESE	745	737	−1.2%
IV1	railing S	606	589	−2.8%
IV1	railing SW	549	552	+0.5%
IV1	railing E	413	404	−2.2%
IV1	garden S	811	797	−1.7%
IV1	tilt 50 SSW	792	784	−1.1%
IV1	tilt 25 ESE	707	695	−1.7%

3. What your home uses, and when

We use the IDEAL Household Energy Dataset: second-by-second electricity readings from 255 UK homes, released by the University of Edinburgh. Each household said in the study's own survey how many days a week someone is home in the daytime. We split them into some days or less (0–4) and most days (5–7). We first tried three groups. But homes answering 0–1 days used no less of a typical kit's output (two panels on a south-west railing in Bristol, 620 kWh a year) than those answering 2–4: 462 and 449 kWh on average, across 5 and 22 homes. A third answer would only have shown noise. Homes occupied most days used more: 498 kWh. 37 homes have at least 300 complete days in one year of readings; a missing day borrows the nearest complete weekday or weekend day within two weeks. We keep 14 of them: in each group, the homes at evenly spaced points from the 10th to the 90th percentile of how much of a typical kit's output they use. Each is averaged to 10 minutes and scaled to the yearly use you give. Your results show the middle home, and the range is the middle 80% of homes like yours.

Then, for every 10 minutes of the year, with each hour's solar output held steady across the six steps around the moment PVGIS measured it (about 20 minutes before to 40 minutes after the hour):

used = Σ min(solar, home)·Δt exported = Σ max(solar − home, 0)·Δt

Averaging a home's use over 10 minutes hides short spikes, such as a kettle, that the panels can't fully cover, so it flatters the share used. We measured that on the 14 homes, against their 1-minute readings and two panels on a south-west railing: 10-minute data puts the share used at home 1.7 percentage points too high on average (0.9 to 3.3), about 3% of the saving. Sunshine varies within the hour too, which works the same way; we haven't measured that, because PVGIS only gives hourly sunshine.

If you're not sure of your yearly use, we offer Ofgem's typical figures for single-rate meters from 1 July 2026: 1,600, 2,500 and 3,800 kWh.

4. A battery

Batteries are outside the plug-in rules, so the switch models one fitted conventionally. Surplus solar charges it and evening demand empties it. The round-trip loss is split evenly between charging and discharging. Its standby draw comes from the battery while it has charge, and from the grid when it is empty.

Defaults: 2 kWh usable, 800 W, 85% round trip and 10 W standby. The source is HTW Berlin's 2025 lab tests of 22 home batteries, where battery efficiency averaged 95.7% before conversion losses (which are largest at night-time loads) and standby draw ranged from 6 W to 60 W. We show payback only once you enter an installed price, because installation costs vary too much to guess.

5. Money

saving a year = (used × unit rate + exported × export rate − battery standby from grid × unit rate) / 100 payback in years = (kit price + stated delivery) / saving a year

Your unit rate starts at Ofgem's price cap for your region (Direct Debit), October to December 2026. We read it from Ofgem's regional tables and refresh it every quarter. Ofgem: no VAT on electricity from 1 October 2026 to 31 March 2027; rates are as billed. Export starts at 0p: Ofgem's Smart Export Guarantee guidance asks for the installation and the installer to be certified (MCS or equivalent), which a kit you plug in yourself isn't. You can change both.

The verdict: Yes if payback is 8 years or less and the saving at least £100 a year. Yes, modestly if payback is 8 years or less with a smaller saving. Borderline up to 12 years. Probably not beyond that. Eight years leaves margin inside typical microinverter warranties of 10–12 years. Paybacks are rounded up and savings compared to the pound, so the figures shown never contradict the verdict.

When none of the kits we list comes with a mount for your spot, there's no real price to divide by. The answer then gives the price that pays back in 8 years instead: 8 × the yearly saving, rounded down to £10.

6. Shade and direction

Shade is your estimate, not geometry. "A little" keeps 75% of direct sunlight and 90% of diffuse light. "A lot" keeps 50% and 80%. If you're not sure of your direction, we work out all eight compass points and show the fifth best (three do worse, four better), with the range. Default tilts are 90° for railings and walls, 35° for a garden stand and 15° for a flat-roof frame. You can change the tilt.

7. Kits

The government's product specification requires every plug-in kit to be listed as compliant on the ENA Type Test Register before it goes on sale. Since 2024 that register runs on ENA Connect Direct, which only installers and manufacturers can search. We list a kit only when its seller prints that register reference. Prices are read on the seller's own site and dated. Delivery is included where the seller states it, and shown as extra where it's charged at checkout.

A price only counts towards payback if it includes a mount for your spot. A garden kit's price says nothing about what a railing bracket costs, so kits without the mount you need are listed separately, with their saving but no payback. Kits are ranked by payback for your home, never by commission, and any affiliate link is labelled. The list covers 6 kits; we add more as sellers show their register entries.

8. What we don't model

Sunshine is hourly, so brief bursts of sun and cloud within an hour are smoothed (see section 3).
Shade from your building, a balcony above or trees beyond your estimate. A balcony is usually a shaded place, so start with "A little" if unsure.
Extra output from the back of bifacial panels, which makes us slightly cautious for those kits.
Panel ageing, future electricity prices and time-of-use tariffs.
A typical year is typical: any real year can be sunnier or duller.

Sources and credits

Sunshine and PV model: PVGIS 5.3, European Commission Joint Research Centre. PVGIS © European Union, 2001–2026. Data fetched 30 October 2026.
Household use: Goddard, N. et al. (2021), IDEAL Household Energy Dataset, University of Edinburgh, CC BY 4.0; Pullinger, M. et al. (2021), Scientific Data 8, 146. We resampled and scaled the readings as described above.
Postcodes: OS Code-Point Open. Contains OS data © Crown copyright and database right 2026. Contains Royal Mail data © Royal Mail copyright and database right 2026. Contains National Statistics data © Crown copyright and database right 2026.
Price cap, regions and typical use: Ofgem and Ofgem's typical consumption decision, Open Government Licence v3.0. Rates checked 30 October 2026.
Batteries: HTW Berlin, Stromspeicher-Inspektion 2025.
Formulas checked against pvlib 0.16.1.
Rules: SI 2026/848, DESNZ, plug-in solar and its Interim Product Specification; Ofgem's Smart Export Guarantee guidance for generators. All read on 30 October 2026.