Running costs are one of the few things in this subject a householder can work out unaided. The arithmetic is trivial; what makes most published figures unreliable is that the numbers fed into it come off the wrong line of a specification sheet, and are multiplied by a tariff rate that expired two quarters ago. What follows is the method, its inputs, and the points at which it stops being trustworthy.
The arithmetic, in one line
Consumption is power multiplied by time, and cost is consumption multiplied by the unit rate.
Electrical input in kW × hours run = kWh consumed. kWh × the unit rate in p/kWh = the running cost.
A wall-mounted split whose published electrical input is 1.08 kW, running four hours on a warm evening, uses 1.08 × 4 = 4.32 kWh. The last multiplication is left to the reader deliberately: the only rate that gives a true answer is the one on the reader’s own bill. The rate is published and the electrical input is on the data sheet. The hours are the hard part, and no web page can supply them — they are a property of the building, not of the machine.
The electricity rate this page uses, and its date
Ofgem’s energy price cap for 1 July to 30 September 2026 sets a domestic electricity unit rate of 26.11p per kWh. Ofgem republishes each quarter, and has already published a separate set of figures for 1 October to 31 December 2026 which it states cannot be compared directly with the current period.
Three qualifications matter more than the number. The cap governs default and standard variable tariffs, so a household on a fixed deal, a prepayment meter or a time-of-use tariff pays something else. It is a Great Britain average across regions. And any p/kWh figure is only as good as its date stamp: an undated rate is arithmetic built on sand.
Why the kW in the name is not the electricity it uses
The commonest error is treating the headline kW as consumption. It is cooling output — heat moved out of the room per unit of time. The electricity drawn to move it is a fraction of that, because the machine is a heat pump, not a heater. Published UK specification data for one current domestic wall-split range makes the gap concrete.
| Rated cooling output | Rated electrical input | EER | SEER | Label kWh per year |
|---|---|---|---|---|
| 2.5 kW | 656 W | 3.81 | 7.0 | 125 |
| 3.5 kW | 1,080 W | 3.24 | 6.6 | 186 |
| 5.0 kW | 1,562 W | 3.20 | 7.0 | 250 |
The 3.5 kW unit a homeowner might assume draws 3.5 kW in fact draws a little over 1 kW at its rated point — so putting the output figure into the arithmetic overstates the running cost by more than threefold.
Part load is the normal condition
Even the rated input overstates ordinary consumption, because an inverter-driven compressor modulates: it varies speed to match the heat actually entering the room instead of cycling on and off at full tilt. Minimum cooling output on those three machines is around 0.89 to 0.9 kW, so the 5.0 kW unit turns down to roughly 18 per cent of its rated output, and draws accordingly.
The rating standard is built around the same behaviour. Assimilated Commission Regulation (EU) No 206/2012 tests every machine against one fixed European reference cooling season of 2,602 hours, within which the full-load 35°C design condition is assumed to occur for just 13. That season is a yardstick for comparing machines, not a description of a West Yorkshire summer, which sits well below it. The point it proves is about the machine: it is designed to spend virtually its whole life throttled far below rated output.
SEER and EER answer different questions
EER is a single-point measurement at 35°C outdoor and 27°C indoor — the machine flat out on the hottest test condition. SEER is weighted across the temperature bins of that reference season and across part-load running, and it legally counts thermostat-off, standby, off and crankcase-heater consumption against the machine as well. Because a unit spends most of its life at part load, where it is more efficient, the seasonal figure comes out well above the instantaneous one — on the machines tabled above, between about 1.8 and 2.2 times the EER. That is a small sample from one product range rather than a market law, but the direction is structural.
For scale: under the same Assimilated Commission Regulation (EU) No 206/2012, the Great Britain legal minimum for a new unit under 6 kW using a refrigerant with a global warming potential above 150 is SEER 4.60 and SCOP 3.80, and under Assimilated Commission Delegated Regulation (EU) No 626/2011 the top A+++ label band starts at SEER 8.50. Building Regulations guidance for England — Approved Document L, Volume 1 — separately says the seasonal energy efficiency ratio of an air conditioner working in cooling mode should be a minimum of 4.0: a should, not a prohibition.
What the annual kWh figure on the label actually is
Data sheets print an annual consumption in kWh. It is reproducible arithmetic, not a measurement:
rated cooling capacity in kW × 350 hours ÷ SEER = the label kWh per year.
2.5 × 350 ÷ 7.0 = 125. 3.5 × 350 ÷ 6.6 = 185.6, printed as 186. 5.0 × 350 ÷ 7.0 = 250 — the exact figures on those three data sheets. The 350 hours are not a survey of anyone’s summer: they are equivalent active-mode cooling hours set by the same regulation so that every machine is compared on identical terms. The label figure is a comparison basis and nothing else — rank two units with it, do not forecast a bill with it. A Yorkshire home will very probably run far less cooling than that.
The solar crossover, stated honestly
Cooling demand and solar generation broadly coincide in season and in time of day, which is genuinely favourable — and routinely overstated. Measured data is the better starting point. Sheffield Solar’s PV_Live series publishes half-hourly output for Britain’s whole solar fleet alongside its installed capacity, which between them give a figure in kWh per kW installed.
| Day in 2026 | Measured GB yield | Equivalent from 4 kWp |
|---|---|---|
| 11 June — dullest day of the summer | 1.414 kWh/kWp | 5.7 kWh |
| Summer average, 1 June to 31 August | 4.008 kWh/kWp | 16.0 kWh |
| 13 August — hottest UK day of the year | 4.582 kWh/kWp | 18.3 kWh |
| 21 June | 4.636 kWh/kWp | 18.5 kWh |
| 19 July | 4.835 kWh/kWp | 19.3 kWh |
| 12 July — best day of the summer | 5.489 kWh/kWp | 22.0 kWh |
A caveat travels with every figure there: PV_Live measures the GB fleet, which includes ground-mounted plant and is weighted towards southern England, so it is the upper end of what a Yorkshire roof does rather than a measurement of one. Those are also gross array figures, before whatever else the house is running takes its share. Modelled for the locality instead, the European Commission’s PVGIS tool puts a 4 kWp array at 35° facing south near Wakefield at roughly 3,850 kWh a year.
Where the crossover breaks down
Timing is the real constraint, not total energy. A day’s generation in kWh is not power available at a moment. Scaled from that measured 21 June day, a 4 kWp array peaked at about 2.05 kW and sat above 1.08 kW for roughly nine hours, from about 08:00 to 16:45 UTC. Live generation alone therefore covers about nine hours of that 1.08 kW unit, and the rest of the day’s output is reachable only through battery storage. The array never reached 3 kW at all, so a 3 kW whole-house cooling load could not have run on live generation at any point that day.
The evening is the problem. On 19 July 2026 the GB fleet peaked at 13:30 UTC, at 56.8 per cent of installed nameplate. By 19:00 BST output was 27.7 per cent of that peak, 11.3 per cent an hour later, 1.1 per cent by 21:00 BST and zero by 22:00. Masonry holds heat, so a bedroom in a Victorian terrace or a 1930s semi is at its worst hours after the roof has finished.
Overcast days are the days people want cooling. The dullest measured day of summer 2026 gave a 4 kWp array 5.7 kWh in total, against 18.5 kWh on 21 June. More than half the light reaching a Yorkshire roof on an average June day is diffuse rather than direct.
The hottest days are not the sunniest days. On 13 August 2026, the hottest UK day of the year at 38.1°C at Kew Gardens, the fleet delivered less than it had on 19 July and about 16 per cent less than on the summer’s best day. British heatwaves often arrive humid and hazy, and modules lose a little efficiency as they get hot. Cooling demand and generation share a season and a time of day; they do not peak on the same afternoon.
And it is not free. Electricity used to cool a house is electricity that would otherwise have been exported. Whether self-consumption beats export turns on the household’s own export tariff — a separately contracted rate, not the same number as the import rate, and a question for a specific bill. What holds without qualification is that a solar array shifts part of a summer cooling load off the meter in the middle of the day, and storage moves the rest into the evening.
The winter half of the same machine
A reversible unit runs the cycle backwards and heats — and on the rating standard’s own weighting that is the larger half: 1,400 equivalent full-load hours for the average heating season against 350 for cooling. Seasonal heating efficiency on that same wall-split range runs SCOP 4.0 to 4.3 — roughly four units of heat for every unit of electricity, where a plug-in electric heater returns one. Capacity holds up in the cold better than most people expect, too: a machine with 3.3 kW nominal heating output is still rated at 2,600 W with −7°C outside.
What cannot be claimed is that a roof pays for it. The same modelled Wakefield array averages about 4 kWh a day in December against about 15 in July, and a more typical 30° south-west-facing roof near Hemsworth about 3.15 kWh: four times as much running, in the season the roof gives least. The winter case is a tariff-and-efficiency argument, not a solar one. Building Regulations guidance for England points the same way, saying reversible heat pump systems should be designed such that they are optimised for heating — one reason the same box is called an air conditioner by one trade and an air-to-air heat pump by another.
What this method cannot tell you
The hours remain the dominant unknown, and they belong to the fabric, not the machine. A west-facing bedroom under a loft conversion, a back room in a solid-walled Victorian terrace and a well-insulated new-build on an old colliery estate do not run the same hours through the same July. That is why Building Regulations guidance in England says a comfort cooling system should be specified from a heat gain calculation for the dwelling, worked with CIBSE Guide A and the manufacturer’s guidance, and should not in most circumstances be sized above 120 per cent of the design cooling load. Sizing and running cost are one question asked twice.
Label figures are laboratory figures, too. Redoing these sums honestly needs three things: the electrical input in kW from the unit’s own data sheet rather than the number in its name, a realistic estimate of hours for that specific room, and today’s unit rate off today’s bill. Everything else here is context for those three.