The Five Numbers You Need Before You Model Solar At All

Most online solar calculators ask for your postcode, your roof direction, and maybe a rough monthly bill. Then they produce a figure. Ten panels, 4,200 kWh a year, £680 saved, payback in 9.4 years. It looks authoritative because it has a decimal point in it.

That figure is almost entirely a function of your postcode. The calculator has modelled the sun landing on a generic roof in your region and multiplied it by an assumed self-consumption rate, usually somewhere between 30% and 50%, because it has no idea what your house actually does with electricity. Change the assumed self-consumption from 35% to 55% and the savings figure moves by hundreds of pounds a year. That single unknown swamps everything else in the model, and it’s the one thing the calculator never asks about.

You can do much better, and you need surprisingly little to do it. There are five numbers. Get all five, and a spreadsheet you build in an afternoon will out-predict any postcode-only tool on the internet. Get four of the five and you’re still ahead. Here they are, what they’re for, and where to actually find them.

1. Annual consumption in kWh

This is the anchor. Everything else is expressed as a fraction of it.

Do not use your bill’s monetary total and divide by a unit rate. Standing charges, VAT, and any residual debit balance will corrupt it. Go to the annual statement your supplier is legally required to send you, or to the usage section of your account, and find the kWh figure. Octopus shows it under Usage as a downloadable CSV. British Gas buries it in the annual summary PDF. If you have a smart meter and you’ve signed up to a data access service, you can pull half-hourly readings, which is better still.

A useful reality check: UK Ofgem’s medium typical domestic consumption value is 2,700 kWh of electricity a year as of the 2025 revision, down from 3,900 kWh several years ago. If your number is 6,500 kWh, you are not a medium household. You probably have a heat pump, an EV, electric hot water, or a home office with a small datacentre in it. That matters enormously, and no postcode calculator will ever know.

Worked example, which I’ll carry through the rest of this piece. Call the household Flat-Roof Semi in Reading:

Annual import (Oct 2024 – Sep 2025):  5,840 kWh
Standing charge:                       excluded
Export currently:                      none (no solar yet)

2. Daytime share

Here’s where the real money lives. Solar generates between roughly 07:00 and 19:00 in summer, and roughly 09:00 to 16:00 in winter. Electricity you use inside that window is worth your full import rate, call it 26p/kWh. Electricity you export is worth your SEG rate, which for most 2025-26 tariffs sits between 4p and 16.5p. The gap between those two numbers is the entire economic case for sizing carefully.

So: what fraction of your annual kWh lands in daylight hours?

If you have half-hourly smart meter data, this is a pivot table. Load the CSV into Excel or Google Sheets, extract the hour from the timestamp, and sum consumption by hour bucket. Free tools that do it for you include Hugo Energy App and the Octopus Compare app on iOS, both of which will read your consumption via the supplier API. For a spreadsheet approach, =SUMIFS(kwh_column, hour_column, ">=8", hour_column, "<16") divided by the annual total gets you most of the way.

Reading semi, actual result:

28% daytime. That’s low, and it tells you something a calculator never would: this household is evening-heavy, probably charging an EV overnight on a cheap rate. A 5 kWp array in Reading generates roughly 4,600 kWh/year. Without a battery, this house self-consumes maybe 1,100 kWh of it and exports 3,500 kWh. At 26p saved and 15p exported, that’s £286 + £525 = £811. Perfectly fine, but the shape of it is 65% export income, not bill reduction, which changes how you think about tariff risk entirely.

Add a 10 kWh battery and daytime share stops being a constraint. Self-consumption jumps to maybe 3,000 kWh, and the sum becomes £780 + £240 = £1,020. Only £209 better for £4,000 of battery. That’s a 19-year payback on the battery alone, and you only discover it because you knew the 28%.

3. Peak demand

Annual kWh tells you volume. Peak demand tells you about the inverter and the battery’s power rating, which is a completely separate question and one people get wrong constantly.

What you want is the highest half-hourly consumption you see in a normal year, converted to kW. A 1.4 kWh half-hour reading means an average draw of 2.8 kW over that period. Look for the 99th percentile rather than the absolute maximum, because one freak reading from the day you ran the oven, the kettle, the immersion heater and a tumble dryer simultaneously shouldn’t drive a £1,500 hardware decision.

Reading semi, half-hourly maxima:
  absolute max        3.9 kWh  → 7.8 kW  (Christmas Day)
  99th percentile     1.6 kWh  → 3.2 kW
  95th percentile     1.1 kWh  → 2.2 kW
  median              0.28 kWh → 0.56 kW

That median of 0.56 kW is the crucial one for battery sizing. A battery rated at 3.68 kW continuous covers the 99th percentile comfortably. Spending extra on a 5 kW or 6 kW unit buys you coverage for about 1% of half-hours, and during those half-hours the grid quietly tops up the difference anyway. Meanwhile the 0.56 kW median means a 10 kWh battery discharging at baseload lasts about 18 hours, so overnight coverage is never the binding constraint. Evening cooking peaks are.

4. Usable roof area

Not roof area. Usable roof area, which is a much smaller number than the one you measure off Google Maps.

Panels in 2026 are typically 1.72m × 1.13m (roughly 1.95 m²) and 440-450 W. So a modern panel delivers about 225 W/m². To get 5 kWp you need about 22 m² of genuinely clear, rectangular, shade-free roof.

What eats into it: 400mm clearance from the ridge, 400mm from the eaves and each verge for fire-service access and wind uplift, plus anything physically in the way. Vent pipes, satellite dishes, a chimney, a dormer, a roof window.

Measure it properly. Open your property on Google Earth Pro (free, desktop) and use the ruler tool on the roof plane, or pull the roof dimensions off your Land Registry title plan if you have one. Better still, get on a ladder with a tape measure and write down the actual rafter-to-rafter span.

Reading semi, south-west facing rear roof:
  gross plane          8.4 m × 4.1 m  = 34.4 m²
  less ridge/eaves setback (0.4 m each)   = 8.4 × 3.3 = 27.7 m²
  less verge setbacks (0.4 m each)        = 7.6 × 3.3 = 25.1 m²
  less soil vent pipe exclusion zone      = 23.2 m²
  panels that fit (portrait, 2 rows of 5) = 10
  array size at 445 W                     = 4.45 kWp

The postcode calculator offered 14 panels. Ten fit. That’s a 29% error in the input that drives every other output, and it appeared before any physics was done.

5. Azimuth and pitch

The last two are a pair, and they’re cheap to obtain. Azimuth is the compass bearing your roof faces, where 180° is due south. Pitch is the angle from horizontal.

For azimuth, stand in the garden with your phone’s compass app pointed perpendicular to the roof ridge, then subtract 90°. Or open the property in Google Maps satellite view, draw a line along the ridge, and read the bearing. For pitch, a phone spirit-level app held against a rafter in the loft works; alternatively, most UK houses built after 1960 sit between 30° and 40°, and 1930s semis are usually 40-45°.

Then feed all of it into PVGIS, the European Commission’s free tool at re.jrc.ec.europa.eu. It’s the same irradiance database most commercial software uses underneath, it accepts latitude, longitude, azimuth, slope, system loss percentage and peak power, and it returns monthly and annual generation. MCS-certified installers use MCS Standard MIS 3002 lookup tables; PVGIS is generally closer to reality because it uses site-specific satellite irradiance rather than regional bands.

PVGIS for the Reading semi at 4.45 kWp, azimuth 225° (south-west), slope 42°, 14% system loss:

Annual:  4,062 kWh   (913 kWh per kWp)
Jun:       512 kWh
Dec:        94 kWh

Note the ratio. December produces 18% of what June does. Any model that treats generation as an annual average and calls it a day will badly overstate winter self-consumption, because winter is exactly when your usage is highest and your generation is a rounding error. This is where the monthly breakdown earns its keep.

What the five numbers together buy you

Stack them and the model becomes a monthly energy balance rather than a single multiplication. For each month you have generation from PVGIS, consumption from your annual figure split by a seasonal shape, and a daytime share telling you how much of that consumption can be met directly. Battery throughput is bounded by usable capacity and your median draw. Export is whatever’s left.

The Reading semi’s honest answer: 4.45 kWp, no battery yet, roughly £790/year against £6,400 installed, a 8.1-year simple payback with the SEG at 15p. The slick calculator said £1,240/year and 5.2 years, because it assumed 14 panels and 45% self-consumption on a house that manages 28%. Both of its key assumptions were wrong in the same optimistic direction, and the errors compounded.

Once you’ve got the five numbers, the next move is turning half-hourly consumption into an hour-by-hour dispatch model, which is where battery sizing stops being guesswork. That’s the ground covered in Sizing From Your Own Consumption Data, and it assumes you already have these five in hand.

One caution on the five. The daytime share you measure today is the daytime share of a household without solar. Install panels and behaviour shifts: the dishwasher moves to 1pm, the immersion heater gets a timer, you start caring what the inverter app says. Measured daytime share is a floor, not a forecast. Build the model on the floor, then treat any behavioural gain as upside you have to earn rather than a number you get to assume.

Go and find your annual kWh first. Everything else follows from having a real denominator.