What a unit of your own electricity actually costs

Per watt prices the hardware, not the electricity

We publish a price index of every approved kit, measured in pounds per watt. The median is £0.82 and the best is £0.45. It is a good measure and we are keeping it, but buried in its own limitations section is a sentence that has been quietly bothering me since I wrote it: per watt is a price measure, not a recommendation.

The problem is that nobody buys watts. A watt is a rating on a label, a statement about the panel under test conditions that Britain almost never provides. What you are actually buying is kilowatt hours, delivered slowly, over years, onto a roof or a railing that faces some particular direction at some particular angle. Two kits at the same price per watt can deliver wildly different amounts of electricity depending on where they end up, and the per watt figure cannot see any of that.

So we have built a second measure. Take the price you actually pay, divide it by the electricity the kit will actually generate over its life, and you get a number you can hold directly against the rate on your bill. It is the same arithmetic the energy industry uses for a wind farm or a gas plant, applied to the smallest generator you can legally own.

What the numbers are

Every figure below is pence per kilowatt hour generated, over twenty years, for a kit bought today in London. The method is at the bottom and you can reproduce all of it.

The spread inside that table is the point. The cheapest kit is not marginally better value than a 400W kit, it is two and a half times better per unit of electricity, and that gap holds whatever you bolt it to. A reader choosing between them on sticker price alone, £432 against £449, would see almost no difference at all.

Generated is not used, and the difference is large

Here is the complication that most people writing about solar savings skip, and it deserves to be near the top rather than in a footnote. A unit your panels make at one o'clock on a Tuesday while the flat is empty is not a unit you have saved. It goes to the grid, and unless you are on an export tariff it earns you nothing at all.

Our calculator models self-consumption from your baseload and your occupancy. For a typical case it lands near 49%, which means roughly half of what you generate is genuinely displacing electricity you would otherwise have bought. That is what turns 2.4p a unit generated into 4.9p a unit used. Both numbers are true and they answer different questions. If you want to know whether the kit is good value as a generator, use the first. If you want to know what it is doing to your bill, use the second.

Even the second number is a sixth of the standard rate. I want to be careful here, because this is the point where solar writing usually starts overselling. Payback still takes years. The saving is real but it is not dramatic in any single month. What the per unit figure tells you is not that you will get rich, it is that you are buying electricity at a fixed price, decided today, for twenty years, while the alternative is repriced every quarter by a regulator.

Overpanelling works in Britain, and our dim skies are the reason

Every compliant kit is capped at 800 VA of export. So why does a 940W kit make sense at all, when 140 of those watts appear to have nowhere to go?

Because the cap bites only at the very top of the curve, and in Britain the curve rarely gets there. Panels are rated under test conditions of full midday sun at 25 degrees. A British afternoon in September delivers a fraction of that. The excess panel capacity is not wasted, it is what drags the kit up to a useful output on the ordinary overcast days that make up most of our year, and it only ever gets trimmed on the handful of bright still afternoons when the inverter hits its ceiling.

I wanted to know how much that trimming costs, because it is the weakest assumption in this whole calculation. So rather than pick a number and hope, I ran the cheapest kit across the full plausible range. With no clipping at all it costs 2.36p a unit. At an aggressive 8% it costs 2.57p. The answer barely moves. That is worth saying plainly: the conclusion does not depend on my assumption, which is the only reason I am comfortable publishing it. In a country with Spanish light the sum would look very different and overpanelling would be a real cost.

Where you put it is a price, and now you can see it

This is the part I think will be most useful, because it makes something visible that has only ever been described in vague terms.

Advice about mounting normally arrives as an adjective. South-facing is "best". A balcony is "fine". East is "not ideal". None of that tells you what it is worth. Running the same kit through PVGIS at each realistic mounting turns those adjectives into money:

A vertical south-facing railing gives up 27% against a properly angled roof. Turn that railing to face east and you are down 49%. That is the real cost of a balcony, and for most renters it is not a choice, it is simply where the building points. Which is exactly why it should be priced rather than waved away: a kit on an east-facing railing at 4.7p a unit is still an excellent purchase, and a reader deserves to know that rather than be told their flat is unsuitable.

It also changes which kit you should buy. On a good roof the gap between a cheap kilowatt and an expensive 400W kit is wide. On a poor mounting it is wider still, because the fixed cost of the kit is spread over fewer units. The worse your position, the more the value per watt matters.

We found a fault in our own calculator doing this

I would rather report this myself than have someone else find it.

Our savings calculator pulls live satellite data from PVGIS, which is the right thing to do. But it was asking for the latitude-optimal tilt every single time, 42 degrees for London, no matter what the panels were actually fixed to. It handled direction correctly. It never asked about angle at all. Which means a kit bolted flat to a balcony railing was being modelled as though it sat on a properly angled roof.

The size of that error is not trivial, and it fell hardest on precisely the readers this site is written for. A south-facing railing was overstated by 37%. An east-facing railing was overstated by 54%. Balcony railings are the defining plug-in solar mounting in this country. We were telling those readers they would generate half as much again as they really will.

It is fixed. Advanced mode on the calculator now carries a mounting input, and the angle it implies is what goes to PVGIS. The default reproduces the old behaviour exactly, so nobody's saved figures have shifted under them without warning, and the changelog carries the whole thing including the dates it was wrong between. The uncomfortable part is that building a better measure is what exposed it. The fault had been there since the calculator first called PVGIS, and a per watt view of the world would never have surfaced it.

How this is calculated

Deliberately dull, and published so that anyone can reproduce or attack it.

• Price is the lowest we can find including VAT and delivery, read on the seller's own page rather than a press release, on 9 September 2026. Where a checkout differs from a listing, the checkout total wins.
• Generation is PVGIS-SARAH3 from the European Commission Joint Research Centre, for London, 1 kWp, 14% system losses, crystalline silicon, free standing, at the tilt and azimuth named in the table above.
• Life is twenty years with 0.5% annual panel degradation, giving an average output factor of 0.954 across the period.
• Clipping against the 800 VA export cap is allowed for on overpanelled kits, and the sensitivity of the result to that allowance is published above rather than hidden.
• Self-consumption, where quoted, is 49%, taken from our own calculator's modelled band for a typical household.
• The grid comparison is the Ofgem price cap for 1 October to 31 December 2026, 26.32p per kWh including VAT, direct debit, England, Scotland and Wales. Off-peak rates are indicative.
• Not included: the cost of money, any mounting hardware you buy separately, inverter replacement, and any export income. All four would move the number, three of them upwards.

What this measure cannot tell you

Four limits, and they matter as much as the figure.

It assumes the kit lasts twenty years. Panels almost certainly will. The microinverter is the part I would expect to replace first, and nothing in this market has a twenty year field record because the market is three weeks old. If an inverter fails in year eight and costs £150 to replace, add roughly half a penny per unit.

It is London. Every figure moves with latitude and local cloud. Glasgow will be meaningfully worse, Plymouth meaningfully better. Run your own postcode rather than trusting the London number.

It says nothing about quality, warranty, or whether the thing turns up. The cheapest kit per unit is a trade product from a builders' merchant, and lead times are tracked separately on where to buy.

It is a comparison against a rate that moves. The 26.32p figure changes on 1 January 2027. The solar figure does not change, which is rather the point, but the size of the gap will.

What I think this means

For three weeks we have been reporting this market in the language of hardware: how many kits are approved, what they cost, what they cost per watt. That was the right place to start because the legal question came first and people needed to know what they could lawfully buy.

The per unit view reframes it. A plug-in solar kit is not a gadget you buy and then hope pays for itself. It is a twenty year fixed-price contract for a slice of your own electricity, at a price you set on the day you buy it, and at current prices that slice costs somewhere between a tenth and a third of what the grid charges, depending on where you can physically put it. That holds on a north London balcony pointing the wrong way. It is a stronger case than the one the industry usually makes for itself, and it is stronger precisely because it is smaller and more specific.

We will publish this figure alongside the pounds per watt on the price index from the next reading, and I will be glad to be corrected on the method. If you think the twenty year assumption is generous, or the clipping allowance is wrong, or the self-consumption band is optimistic, tell me at [email protected] and I will publish the correction.

Sources: PVGIS-SARAH3, European Commission Joint Research Centre, queried 15 September 2026 for London at the tilts and azimuths stated. ENA G98/G99 Type Test Verification Register, plug-in solar device type, checked 9 September 2026. Seller pages read on the same date. Ofgem price cap for 1 October to 31 December 2026. Plug Solar Hub takes no commission, fee or discount code from any seller of a compliant kit, and there are no affiliate links on this page. Educational information, not financial or electrical advice.