Energy dataLast updated: 418 monitored systems, twelve months to date

What Heat Pumps Actually Achieve in Real British Homes

Contents

Almost everything written about heat pump efficiency is either a manufacturer’s laboratory rating or somebody’s anecdote. This page is neither. It is what several hundred British households have measured in their own homes over a year and chosen to publish.

What the number means

A heat pump’s coefficient of performance is how many units of heat it delivers for each unit of electricity it uses. A COP of 4 means four kilowatt-hours of heat from one kilowatt-hour of electricity — which sounds impossible until you remember the machine is moving heat that was already outside. COP is not the same measure as combustion efficiency.

That ratio is one input to a running-cost comparison, alongside the applicable electricity and gas prices and what the old heating system used.

What these homes actually achieve

Grouped by what the meters actually covered, because that changes the answer more than anything about the machines does.

Whole system metered

The fullest arrangement: what went in and what came out, both measured.

Everything together: 3.99middle half 3.59–4.29348 systems
Heating the rooms: 4.25middle half 3.89–4.66234 systems
Heating the hot water: 3.09middle half 2.68–3.44216 systems

Most of the system metered

Closer to complete, but still not everything the system draws.

Everything together: 3.68middle half 3.29–4.2249 systems
Heating the rooms: 4.15middle half 3.75–4.5428 systems
Heating the hot water: 3.13middle half 2.84–3.5027 systems

Partial metering

Some electricity the system uses falls outside what was measured, so these figures read higher than the whole system would.

Everything together: 3.47middle half 3.27–4.0720 systems
Heating the rooms: 3.71middle half 3.52–4.1513 systems
Heating the hot water: 2.99middle half 2.49–3.2712 systems

The colder the water, the better it does

Median across everything, for whole-system-metered installations only. Grouping by flow temperature across different metering arrangements would confuse the two effects, so this is one boundary.

Flow temperature 35 to 45°C: 3.8950 systems
Flow temperature Under 35°C: 4.0696 systems

Almost all of this is air source

343 air source
4 ground source
1 water source

The sample is almost entirely air source. Within the one metering boundary these figures may be compared in, there are four ground source systems and one water source system, and their results span a wider range than the difference anybody is asking about. No efficiency figure is published by heat pump type, and the counts here are given so the shape of the sample is visible rather than assumed.

The manufacturer ranges overlap

Air source only, whole-system-metered only, and only makes with at least 10 systems. 3 of the 7 makes below vary more between their own installations than the makes vary between each other.

Vaillant182 systems3.78 to 4.414.05 median
Daikin42 systems3.46 to 4.043.79 median
Viessmann29 systems3.84 to 4.294.17 median
Samsung19 systems3.28 to 3.853.56 median
Mitsubishi16 systems3.34 to 4.043.79 median
Grant14 systems3.46 to 4.373.95 median
Panasonic12 systems3.61 to 4.033.79 median

These distributions overlap and do not isolate a manufacturer effect from differences between homes, system design, installation or settings. This is not a ranking and a position in it is not a recommendation.

805 systems publish their monitoring; 418 are used in the figures above. 52 are excluded because the publisher flags something about how they were measured, and rows failing a quality check are excluded too. These are households who chose to publish their own data, not a representative sample of British heat pumps.

Why there is no single figure here

You will have noticed there is no headline number, and that is deliberate. The monitored systems did not all measure the same thing. Some metered the whole system — every unit of electricity in, every unit of heat out. Others metered most of it, and some only part.

A COP that leaves out some of the electricity a system draws reads higher than one that counts all of it, so the arrangements are not comparable. Compare the three groups above and the gap between them is wider than most of the differences anyone would want to draw from this data. Averaging across them would produce the tidy single figure everybody wants, and it would be an artefact of how the meters were fitted rather than a fact about heat pumps.

What happened in a separate volunteer trial

Everything above comes from households who bought a heat pump and then chose to monitor it. What follows is a different kind of evidence entirely: a government-funded demonstration with a common metering specification. Its participants still volunteered, so it is not a representative sample of British homes.

The Electrification of Heat Demonstration Project was funded by the Department for Energy Security and Net Zero and delivered by Energy Systems Catapult. It installed 742 heat pumps free of charge into homes recruited to fill a spread of property archetypes across south east England, north east England and central Scotland, then metered them to a specification the project set rather than the householder. 638 of those homes returned a complete metered year.

It is a trial rather than the market — participants volunteered, the equipment was free, and the designers knew the homes were being measured. Its figures sit on their own axis below, and they are not comparable with the monitored figures above: different households, different years, and meters that covered different things.

Electrification of Heat Demonstration Project

Electrification of Heat Demonstration Project: Heat Pump Performance Cleansed Data: Daily Performance Dataset, 2020-2023. Energy Systems Catapult, edition 1, 19 December 2024. UK Data Service SN 9210.

2.74

units of heat for each unit of electricity, over a whole metered year, in the middle home of 478 air source installations. The middle half of those homes fell between 2.46 and 3.03.

What that ratio is, exactly

Heat counted: Heat measured by the heat pump’s own heat meter (Heat_Pump_Energy_Output). Heat from an immersion or back-up heater is not measured and is not included.
Electricity counted: Electricity measured on the primary wiring of the whole heat pump system (Whole_System_Energy_Consumed), including circulation pumps, immersion heater and back-up heater.

This is not a seasonal performance factor. It is not SPFH2, SPFH3 or SPFH4, and it should not be compared with a figure quoted as any of them. It is conservative for any home that used its immersion or back-up heater: that electricity is counted and the heat it produced is not measured.

It gets worse in the cold, and worse in the warmth too

Every usable day in the trial, sorted by the average outside temperature that day and pooled across all 737 homes with usable readings. Efficiency peaks in the mild middle and falls at both ends — the fall at the top is real and is about short cycling and hot water, not about the weather being hard work.

Heat delivered per unit of electricity by outside temperature, in the Electrification of Heat Demonstration Project
Outside temperature	Heat per unit of electricity	Homes	Days
-5°C to -2.5°C	2.24	503	2,069
-2.5°C to 0°C	2.35	658	6,433
0°C to 2.5°C	2.56	732	25,618
2.5°C to 5°C	2.74	733	49,904
5°C to 7.5°C	2.94	733	68,825
7.5°C to 10°C	3.04	733	74,648
10°C to 12.5°C	2.95	734	76,597
12.5°C to 15°C	2.55	730	83,006
15°C to 17.5°C	2.23	730	61,486
17.5°C to 20°C	1.95	720	22,674
20°C to 22.5°C	1.75	677	3,882
22.5°C and above	1.72	517	1,557

No published cell rests on fewer than 5 homes or 100 days; nothing was suppressed on this build. The coldest and warmest bands are open-ended, so the coldest band also holds the few days below its own label.

Every system type, including the hybrids

Homes with a gas boiler alongside the heat pump are reported separately and never in the headline. Their electricity meter cannot see the gas, and the ratio for a hybrid is a statement about how the two were sequenced rather than about the heat pump.

Whole-year heat per unit of electricity by system type
System	Middle home	Middle half	Homes
Every system in the trialAir source, ground source and hybrids pooled. Shown for completeness; the hybrids in it make it lower than a heat-pump-only figure.	2.70	2.37–3.04	638
Air source heat pumpThe figure most readers are asking about, and the one this page leads with.	2.74	2.46–3.03	478
Ground source heat pumpA small group. Read the middle half rather than the midpoint.	2.75	2.52–3.58	32
Hybrid (heat pump with a gas boiler)A gas boiler shares the heating. The electricity meter cannot see the gas, so this measures how the two were sequenced rather than how the heat pump performs.	2.46	1.91–3.03	128

A home counts towards these figures only with a full metered year: A home contributes a whole-year ratio only with at least 330 usable consecutive-day pairs touching all 12 calendar months. Of 742 heat pumps installed, 739 are in the daily dataset and 638 cleared that bar. Readings run from 27 October 2020 to 28 September 2023.

What this evidence cannot tell you

This is a trial, not the market: A government-funded demonstration project, not the market. 742 heat pumps were installed free of charge by three contractors into homes recruited to fill a spread of property archetypes. Nothing here is a rate for Great Britain or an average British heat pump.
The households volunteered: A volunteer sample, as the depositor records it. Participants applied to take part, most commonly citing sustainability, and the installation was free — so this measures homes whose occupants wanted a heat pump and let it be metered.
How many homes each figure rests on: The daily dataset covers 739 of the 742 installations. Figures split by system type rest on fewer homes still, and every published cell states how many.
Where the trial ran, and why you cannot break these down by area: No geography is stored, and none was fetched. The study declares "No spatial unit". The trial ran in south east England (excluding London), north east England and central Scotland; that is the trial’s footprint, not a geography these figures can be broken down by.
When it was measured: The metered window runs from sample.firstDate to sample.lastDate — before, through and after the energy crisis. The equipment is 2020-22 vintage and the installations were designed to the standards of the day. Note that the catalogue declares the collection as starting on 1 November 2020 while the daily file carries readings from 26 October 2020; the file wins, and the difference is recorded rather than reconciled.
What the meters covered: Efficiency here is the heat pump’s own heat meter divided by the whole-system electricity meter. It is not SPFH2, SPFH3 or SPFH4, and it is conservative for homes that used an immersion or back-up heater, whose electricity it counts and whose heat it cannot see.
Hybrids are reported separately: Homes with a gas boiler alongside the heat pump are reported separately and never in the headline. Their electricity meter cannot see the gas, and the ratio for a hybrid is a statement about how the two were sequenced rather than about the heat pump.
The readings are odometers, not daily totals: The source columns are cumulative meter readings, not daily totals. Every figure here is built from differences between consecutive days; an average of the published columns would be half an odometer reading.
What these figures must not be put beside: Do not place these figures on one axis with the IDEAL household archetypes or with community heat-pump monitoring. Different populations, different metering boundaries, different years.

Energy Systems Catapult (2024), Electrification of Heat Demonstration Project: Heat Pump Performance Cleansed Data: Daily Performance Dataset, 2020-2023 [data collection], UK Data Service, SN 9210, DOI 10.5255/UKDA-SN-9210-1. Contains public sector information licensed under the Open Government Licence v3.0. Published by the UK Data Service under the Open Government Licence v3.0.

Does it still work when it is cold?

This is the question people actually ask, and the trial answers it with a shape rather than a number. Efficiency peaked at 3.04 units of heat per unit of electricity between 7.5°C to 10°C, and fell away on either side of that: 2.24 in the coldest band the trial recorded, and 1.72 at 22.5°C and above.

The coldest bin still records heat output; its ratio is lower than the mild-bin peak. The homes in the trial were using around 29 kilowatt-hours of electricity a day in that coldest band against 10 in the peak-ratio band. These are pooled observational bins, not an estimate of the effect on one particular home.

The ratio also falls at the warm end while mean daily electricity is small. This aggregate does not split each temperature bin into room heating, hot water, circulation and standby, so it cannot establish which of those caused the fall.

Hot water is where they struggle

In the current self-selected, whole-system-metered sample, room heating has a median COP of 4.25across 234 systems. Hot water has a median of 3.09 across 216 systems.

Hot-water production generally uses a higher water temperature than room heating. The monitored distributions show an association, but this read does not isolate temperature, cylinder, schedule or system design as the cause of the gap.

What the flow-temperature groups show

Flow temperature is the water temperature the system sends round the house. In the current whole-system-metered sample, the Under 35°C group has a median COP of 4.06 across 96 systems, compared with 3.89 across 50 systems in the 35 to 45°C group.

The groups contain different homes and systems, so the difference is an observed association rather than a controlled estimate of what changing one home’s setting would do.

What the electricity itself costs is the other half of the sum, and examples of dedicated heat-pump tariffs include Cosy Octopus and EDF Heat Pump Tracker. If a grant is part of your arithmetic, see the current Boiler Upgrade Scheme terms.

What one costs to install

Neither the monitored fleet nor the trial says anything about price. The domestic Renewable Heat Incentive did ask: every applicant reported what their installation had cost. The listed capacity-band cells contain about 84,840contributing applications across heat pumps, biomass and solar thermal; 71,360 are air- or ground-source heat-pump applications. The final statistics cover Great Britain between April 2014 and March 2024.

Two things to hold on to before reading them. They are self-reported and were never audited — they are what households said they paid, not a price anyone checked. And they are spread across a decade in the money of the year each one was paid, with no adjustment for inflation, so there is no single “cost today” in this table and we do not manufacture one.

Domestic Renewable Heat Incentive

Non-Domestic and Domestic Renewable Heat Incentive (RHI) monthly deployment data. Department for Energy Security and Net Zero, from Ofgem data. Final edition, March 2024.

What 84,840 contributing applications reported as installation cost, in Great Britain, between April 2014 to March 2024. These are prices from across that whole decade, in the money of the year they were paid, and they have not been adjusted for inflation. Read them as a record of what people paid, not as a quote for today.

Reported installation cost by technology and capacity band, domestic Renewable Heat Incentive, April 2014 to March 2024, nominal pounds
Size band (kW)	Typical cost	Per kW installed	Installations
Air source heat pump
Less than 5	£7,500middle half £5,500 to £8,590	£1,500middle half £1,200 to £1,830	9,050
6 - 10	£9,400middle half £7,320 to £12,240	£1,190middle half £930 to £1,530	24,980
11 - 15	£12,500middle half £10,000 to £15,000	£1,000middle half £810 to £1,220	17,480
16 - 20	£14,000middle half £11,000 to £16,740	£860middle half £680 to £1,010	7,400
21 - 25	£19,000middle half £15,730 to £24,390	£820middle half £690 to £1,050	690
26 - 30	£21,940middle half £18,000 to £27,440	£800middle half £640 to £1,010	270
31 - 35	£21,650middle half £17,000 to £27,000	£670middle half £530 to £840	260
36 - 40	£29,230middle half £20,000 to £33,910	£800middle half £550 to £940	20
41 - 45	£26,580middle half £19,910 to £32,600	£600middle half £460 to £750	30
Biomass
Less than 5	Not published — typical cost, Biomass, Less than 5 kW. No installations in this band, so there is no cost to reportmiddle half not published — No installations in this band, so there is no cost to report	Not published — cost per kW, Biomass, Less than 5 kW. No installations in this band, so there is no cost to reportmiddle half not published — No installations in this band, so there is no cost to report	0
6 - 10	£13,000middle half £10,000 to £16,100	£1,360middle half £1,000 to £1,700	120
11 - 15	£12,000middle half £8,930 to £16,000	£840middle half £630 to £1,110	1,250
16 - 20	£12,790middle half £9,070 to £16,600	£690middle half £500 to £880	1,350
21 - 25	£14,260middle half £10,000 to £18,000	£600middle half £430 to £720	2,250
26 - 30	£16,440middle half £13,000 to £20,000	£600middle half £470 to £730	1,520
31 - 35	£18,000middle half £17,340 to £23,310	£510middle half £510 to £690	1,300
36 - 40	£19,250middle half £15,000 to £25,000	£500middle half £390 to £660	820
41 - 45	£25,780middle half £18,490 to £34,000	£580middle half £420 to £760	620
Ground source heat pump
Less than 5	£8,590middle half £7,210 to £13,490	£1,800middle half not published — The publisher printed zero here, which is an absent figure rather than a price of nothing	350
6 - 10	£15,000middle half £10,900 to £20,000	£1,890middle half £1,000 to £2,500	3,090
11 - 15	£20,000middle half £15,000 to £27,660	£1,660middle half £630 to £2,230	4,550
16 - 20	£26,860middle half £20,000 to £35,000	£1,520middle half £500 to £1,960	1,000
21 - 25	£30,870middle half £24,920 to £40,290	£1,360middle half £430 to £1,820	1,300
26 - 30	£40,060middle half £30,000 to £55,510	£1,430middle half £470 to £1,970	450
31 - 35	£49,880middle half £35,040 to £71,070	£1,460middle half £510 to £2,140	170
36 - 40	£45,490middle half £32,670 to £67,370	£1,180middle half £390 to £1,730	120
41 - 45	£50,000middle half £38,750 to £75,000	£1,140middle half £420 to £1,720	150
Solar thermal
Less than 5	£4,500middle half £3,450 to £5,990	£1,780middle half not published — The publisher printed zero here, which is an absent figure rather than a price of nothing	4,080
6 - 10	£6,000middle half £4,500 to £8,430	£920middle half £1,000 to £1,240	150
11 - 15	£4,990middle half £4,130 to £7,530	£380middle half £630 to £640	20
16 - 20	Not published — typical cost, Solar thermal, 16 - 20 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Not published — cost per kW, Solar thermal, 16 - 20 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Withheld
26 - 30	Not published — typical cost, Solar thermal, 26 - 30 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Not published — cost per kW, Solar thermal, 26 - 30 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Withheld
31 - 35	Not published — typical cost, Solar thermal, 31 - 35 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Not published — cost per kW, Solar thermal, 31 - 35 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Withheld
36 - 40	Not published — typical cost, Solar thermal, 36 - 40 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Not published — cost per kW, Solar thermal, 36 - 40 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Withheld
41 - 45	Not published — typical cost, Solar thermal, 41 - 45 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Not published — cost per kW, Solar thermal, 41 - 45 kW. Withheld: fewer than 20 installations in this bandmiddle half not published — Withheld: fewer than 20 installations in this band	Withheld

There is no total row, because the size bands do not join up: Less than 5 is followed by 6 - 10, so a 5.5 kW installation is in neither and the bands cannot be added. Bands with fewer than 20 installations are withheld by the publisher. Where the publisher printed a zero we show what it means rather than the number, because a zero here is an absent figure and never a price.

Before you use any of these figures

The prices are self-reported and were never audited: Self-reported by applicants on the RHI application form and not validated by Ofgem. DESNZ removed installations with a zero capacity or cost and any above 45 kW, but did not audit prices.
Nominal money across a decade, not today’s prices: Nominal pounds spanning April 2014 to March 2024, undeflated. There is no deflator in the release, and these figures must not be rebased into today’s money.
A published zero is a missing figure, never a price of nothing: A published zero is an absent figure, never a price. Where a technology and capacity band has no installations the workbook writes 0 rather than a suppression marker.
The size bands do not join up and cannot be added together: The capacity bands are the publisher’s and do not tile: “Less than 5” is followed by “6 - 10”, so nothing between 5 and 6 kW appears in either, and the bands are not a partition that can be totalled.
Great Britain, and a scheme that has closed: Great Britain only, and a closed scheme: the domestic RHI stopped taking applications on 31 March 2022 and was replaced by the Boiler Upgrade Scheme.
This edition replaces its predecessor rather than extending it: The register revises downward, so this edition supersedes rather than extends its predecessor. Earlier editions disagree with it and must not be spliced onto it.
The scheme’s efficiency figures are design values and are not shown here: Seasonal performance factors here are design values from the MCS certificate, truncated at the scheme’s own 2.5 eligibility floor. They are not monitored performance and must never sit on the same axis as a measured figure.

Contains public sector information from the Department for Energy Security and Net Zero (RHI deployment statistics, Ofgem data), licensed under the Open Government Licence v3.0. Licensed under the UK Open Government Licence (OGL).

Whose homes these are

These are households who monitor their own heat pumps and publish the results. That is the strength of the data and its limitation in the same sentence: it is genuine measurement over a year rather than a rating from a test rig, and it selects for people interested enough to instrument their heating.

So read these as what a heat pump can do in a British house, not what an average installation does. This self-selected read cannot quantify the gap or attribute it to equipment, specification, commissioning or household behaviour.

Systems the publisher flags as having a measurement problem are excluded rather than quietly included, as are readings that fail a plausibility check. No household location is stored anywhere in this data.

Frequently asked questions

What is a COP?

The coefficient of performance: units of heat delivered for each unit of electricity used. A COP of 4 means four kilowatt-hours of heat from one kilowatt-hour of electricity. It can exceed one because the delivered heat includes energy moved from outside; it is not the same measure as combustion efficiency.

Why is there no overall average on this page?

Because the monitored systems did not all measure the same thing. Some metered the whole system, others only part of it, and a COP that excludes some of the electricity a system draws reads higher than one that includes it. Averaging across those arrangements produces the single number everyone wants and it does not mean anything, so the figures stay grouped by what was actually measured.

Do these figures apply to my house?

Not directly. These are households who chose to publish their monitoring, which selects for people who are interested in getting it right — well-specified installations, often carefully commissioned. Read them as what a heat pump can achieve in a British home rather than what an average installation does.

Why is hot water so much worse than heating?

In the current self-selected, whole-system-metered sample, median room-heating COP is 4.25 across 234 systems and hot-water COP is 3.09 across 216. Hot-water production generally uses a higher water temperature, but this observational read does not isolate the cause of the gap.

Do heat pumps stop working in cold weather?

No. In the volunteer Electrification of Heat trial's daily dataset, the measured ratio was 2.24 in -5°C to -2.5°C and peaked at 3.04 in 7.5°C to 10°C. This is the heat pump's metered heat divided by whole-system electricity, not a market-wide efficiency rate.

What does a heat pump cost to install?

The final domestic Renewable Heat Incentive table has contributing cost counts for about 71,360 air- and ground-source heat-pump applications across its listed capacity bands. For air source, the Less than 5 kW band has a £7,500 median and the most populated band, 6 - 10 kW, has a £9,400 median. These are unaudited, self-reported nominal prices spanning April 2014 to March 2024, not a current quote.

What flow temperature should I aim for?

In the current self-selected, whole-system-metered sample, the Under 35°C group has a median COP of 4.06 across 96 systems, against 3.89 across 50 systems in the 35 to 45°C group. That is an observed association, not a controlled estimate of what changing one home's setting will do.