Your heat pump's efficiency more than halves in the cold, and the maker publishes the number

Our piece on running an air conditioner as a heater ended with an admission. It said this site did not know what efficiency these units actually achieve in heating mode, because no reversible unit in our database published a seasonal figure we had verified against the manufacturer.

That was true in September. It is not true now, because the figures exist and they are more interesting than a single number would have been.

Mitsubishi publishes the efficiency of its units at six different outdoor temperatures. For the MSZ-AP25VG the top of that range is 7.0 and the bottom is 2.24.

The number that matters is not one number

Here is Mitsubishi’s own declared data for the MSZ-AP25VG, from its Ecodesign product information sheet, for the average heating season. COPd is the declared coefficient of performance at that outdoor temperature: heat out divided by electricity in.

A unit delivering seven units of heat per unit of electricity on a mild October afternoon delivers three on a cold January night. That is not a fault and it is not a manufacturer being evasive. It is the physics: the machine is moving heat from outside air, and when the outside air is colder there is less heat in it and more work to move it.

What it means practically is that the efficiency figure quoted on a product page is a single answer to a question that has six.

Why the headline figure looks better than this

The same document gives the MSZ-AP25VG a SCOP of 4.7 for an average heating season and 5.8 for a warmer one. Mitsubishi’s UK databook gives the MSZ-AP25VGK a SCOP of 4.80. (The tenth of a point between the two is the sort of discrepancy you get between a fiche and a databook revision, and we are not going to pretend it away.)

Either way, roughly 4.8 is the number a buyer sees, and it is genuinely higher than any of the cold-weather figures in the table above except the two mildest.

That is because SCOP is seasonal. It is calculated under BS EN 14825 across a whole modelled heating season, weighting each outdoor temperature by how many hours a year it occurs. Most heating hours in a temperate climate happen between about 2 and 10 degrees, where the unit is efficient, and very few happen below minus 5. The average is dominated by the mild hours.

So SCOP is not a lie. It is the right number for estimating an annual bill and the wrong number for working out what happens on the coldest night of the year. Both questions get asked, usually by the same person, usually in the same conversation.

The two capacities Mitsubishi publishes

There is a second thing in the MSZ-AP UK databook worth pulling out, because it is easy to miss and it changes sizing.

Every model in the range has two heating capacities listed, a nominal one and a UK one, and the UK figure is lower on every single model. These are the three units of this range in our unit database:

For the AP25VGK that is 3.2 kW against 2.64 kW, a gap of about 18%. Across the range the UK figure runs consistently below the nominal one.

If you size a unit from the big number and the real delivered output in a British winter is the smaller one, you have bought a unit that is nearly a fifth short of what you planned for, at the time of year you planned it for. The databook gives you both figures. It is worth reading the row you need rather than the row that comes first.

Where the break-even actually sits

Our heating-mode piece worked out the threshold. At the October 2026 capped rates, a condensing boiler running at 90% efficiency delivers useful heat at about 8.86p per kWh, so an electric unit needs an efficiency of roughly 3 to match it on cost.

Put that against the declared table and the answer falls out:

• At 7 degrees outside, COP 6.2. The unit delivers heat at around a third of the boiler’s cost.
• At 2 degrees outside, COP 4.8. Still comfortably cheaper than gas.
• At minus 7 degrees, COP 3.0. Level pegging. This is the break-even night.
• Below that, gas is cheaper, and the gap widens as it gets colder.

Which is the uncomfortable shape of it. The unit is cheapest to run when you need it least, and it reaches parity with gas on exactly the nights you need it most. That is not an argument against the technology, but it is an argument against assuming a single annual figure describes your worst week.

You can run your own version of this against your own tariff with the running cost calculator, and the unit database records which models are reversible at all. Three of the units we hold are cooling only and no setting will make them heat.

What to do with this

Three things follow.

Size for the cold, not the average. Use the UK heating capacity, and if the unit is your only heat source, work out what it has to deliver on a design-cold day rather than on a typical one.

Expect the bill to be seasonal in shape, not flat. A heat pump’s running cost is not a fixed multiple of a boiler’s. It is a good deal cheaper for most of the winter and roughly level for a small part of it.

Treat the coldest nights as a separate problem. The bivalent temperature in Mitsubishi’s data for this unit is minus 10 degrees, which is the point at which the design assumes supplementary heat. For most of the UK that temperature arrives rarely or never, but “rarely” is not “never”, and a house with no second heat source has nowhere to go on the night it does.

Our piece on when to put the heating on has the dates these temperatures actually arrive across the 25 towns we hold records for, which is the other half of this question.

The short version

The efficiency figure on the box is a seasonal average, and the real figure moves between 7.0 and 2.24 depending on what the weather is doing. Mitsubishi publishes all six points and nobody reads them.

For most of a British winter that is good news: these units are very efficient between 2 and 10 degrees, which is where most of our heating hours sit. The number to plan around is the cold one, and the sizing figure to plan around is the UK capacity rather than the nominal one.

Both are published. They are just not the numbers in the advert.