Self-consumption: the number that decides whether solar pays

The six-to-one ratio

Everything on this page follows from one arithmetic fact.

So generation is not the product. Generation multiplied by what each unit is worth is the product, and self-consumption is what sets the second term.

What the UK’s official method says

MCS publishes the authoritative UK method in guidance note MGD 003, and every MCS installer is required by MIS 3002 to give you a self-consumption figure derived from it, in writing, before you buy.

The method takes three inputs:

1. Annual generation from the array, in kWh.
2. Annual electricity consumption of the property, in kWh, from actual meter readings or bills. Where it cannot be established, MGD 003 directs the installer to use 3,500 kWh.
3. An occupancy archetype, chosen by asking the occupier which best describes them.

The archetypes are the interesting part, because they are the whole behavioural model reduced to three options:

Where occupancy is unknown, the guidance says to use the “in half the day” table. The number of occupants is not an input: MGD 003 treats it as already represented by the annual consumption figure.

You then read a percentage out of a lookup table and multiply it by generation.

The figures, for the most common case

Here is the case a great many British households are actually in: Ofgem’s typical medium consumption of 2,500 kWh a year, and a 4 kWp south-facing array, which on our own modelling generates about 3,676 kWh in London.

That puts you in MGD 003’s 2,500–2,999 kWh consumption tables at the 3,600–3,899 kWh generation row. The PV-only column gives:

Between one fifth and one seventh. Not half, and not the two thirds that a payback figure on an installer’s website often quietly assumes.

Two patterns worth knowing, both visible in the tables:

It falls steeply as the array grows relative to demand. For the same home-all-day household, MGD 003 gives 86% for a tiny array under 300 kWh, 29% at around 2,550 kWh of generation, 23% at 3,750 kWh and 17% at 5,850 kWh. This is the ceiling effect: surplus has nowhere to go but the grid.

Occupancy moves it by a factor of about 1.6. Being home all day rather than out all day raises self-consumption by around 60% in relative terms — real, and smaller than the difference a modestly sized battery makes.

What a battery does

On MCS’s figures, quite a lot. For the same household and the same 3,676 kWh array, the home-all-day row moves from 23% with no storage to 37% at 1.1–2.1 kWh of usable capacity, 45% at 2.1–3.1 kWh and 50% at 3.1–4.1 kWh. Those are the column bands MGD 003 uses, and the capacity that counts is usable rather than nameplate.

MGD 003 also sets a hard ceiling: “the self-consumption cannot exceed 95% of the total annual generation”, however much storage is fitted.

The limits MCS states about its own method

This is the part worth reading before treating any figure as a prediction, and MGD 003 is admirably direct about it.

It is an average, not a forecast. The guidance says self-consumption determined by it “is not a performance prediction for an individual property but rather it is the average self-consumption for a sample of domestic properties with similar occupancies, electricity consumption and solar PV systems.”

Whole categories of load are excluded. MGD 003 says that additional self-consumption from “non-typical domestic loads such as electric space heating, swimming pools, heat pumps, electricity power diverters, electric water heating and electric vehicles is not accounted for in the method.”

That exclusion is a large gap in 2026, and it cuts in the direction that favours solar. A household charging a car at home in daylight, or running a heat pump, will self-consume more than these tables show, so for those households the figure should be read as conservative. How much more, MGD 003 does not say — and neither will we, because we have no source for it.

It is year one only. It does not account for panel degradation, battery degradation, or people changing their habits.

The research behind it predates 2022. It rests on Loughborough University occupancy research, battery modelling developed with Advance Further Energy, and validation by the BRE National Solar Centre.

The correction we made to our own calculators

We had this wrong, so it goes here rather than in a changelog nobody reads.

Both our calculators asked for a self-consumption range and defaulted to one. The rooftop calculator defaulted to 25–60% and the plug-in calculator to 50–90%, both labelled as our assumption rather than a measurement — which was honest about the type of the number but not about its size. Having now read MGD 003 properly, those defaults were too high, and the rooftop one was comprehensively so: the bottom of our range sat above the top of what the official method gives for the same scenario.

So we have changed them:

The direction of the error is the one that should embarrass us: it flattered solar, which is the same failure we criticise elsewhere on this site. The cause was ordinary — we set a plausible-looking range early and did not go back to the primary document until we wrote this page. That is precisely why the site’s rules say figures must come from sources rather than from judgement.

Both remain editable inputs, because the right value depends on your consumption, your generation and your occupancy, and only you know the third.

How to raise your own self-consumption

Ordered roughly by how much they achieve for what they cost.

Move flexible loads into daylight. Dishwasher, washing machine, tumble dryer, immersion heater, car charging. Free, and it acts on the largest single controllable chunk of household demand.

Get your own half-hourly data. Your supplier holds it and you can ask for it. It is the only thing that turns this question from a lookup table into a measurement.

Consider a solar diverter before a battery. Diverting surplus to hot water is cheaper than storing it in a battery, and MGD 003 excludes diverters from its model, so the tables understate what one does.

Then consider a battery. It is the largest single lever and the most expensive one. On MCS’s figures, 3–4 kWh of usable capacity roughly doubles self-consumption for a typical household.

Size the array to the consumption, not the roof. If the goal is value per pound rather than maximum generation, an array that overshoots annual demand is buying kilowatt hours at 26p of value and selling them at 4p.

Get an export tariff regardless. Whatever your self-consumption, the surplus is worth more than nothing. Our SEG table lists every rate we could find, and the gap between the best untied tariff and a common one is roughly threefold.