The December heat pump: why one SCOP figure hides the winter

Worked example: 1960s semi-detached house, 4.93 kW design load. All three figures below come from the same model run on the same house. The annual space heating SCOP is 3.87 in both the flat and the bin-based version; only its distribution through the year differs.

Declared seasonal SCOP

3.87

Space heating, 45 °C design flow. Correct for the year as a whole.

January, with hot water

3.34

What the whole system averages in its most expensive month.

On a −3 °C day

2.64

46% more electricity per unit of heat than the seasonal figure implies.

A right answer to the wrong question

Almost every heat pump running cost estimate in the UK is built the same way. Take the annual heat demand, divide it by a seasonal SCOP, multiply by a unit rate. The SCOP usually comes from the manufacturer's declaration under BS EN 14825, which is a genuinely seasonal figure: it already averages the unit's performance across a reference heating season, including weather compensation and part-load operation.¹

For the annual total, that is fine. The trouble starts when the same figure is used for a month, a day or an hour. A heat pump's efficiency moves a long way with the weather. It works hardest, at the highest flow temperature and the biggest temperature lift, on exactly the days when the house needs the most heat. Apply one flat SCOP to every month and you quietly move electricity out of winter and into the shoulder seasons.

That has consequences people feel:

• The December bill. A householder told to expect a certain running cost sees a winter bill well above the monthly average and concludes the system is failing. Most of that jump is heat demand, which any estimate should show. Part of it is efficiency, which a flat SCOP hides.
• Tariff choice. Time-of-use tariffs reward electricity used in a few night hours. Whether a heat pump can use them well depends on when its electricity falls, and at what efficiency.
• Storage sizing. A thermal store or a home battery is sized for a day, usually a cold one. An annual average says very little about a cold day.
• Oversizing. An oversized unit shows up as cycling in mild weather, which a flat figure cannot see at all.

What we changed

HeatLens now splits the year into one-degree outdoor temperature bins for each calendar month, and works out heat, efficiency and electricity for every bin. The approach follows the bin method in BS EN 14825 and EU Regulation 813/2013, with UK weather and UK operating behaviour in place of the standard's reference climate.¹ ² There is one firm rule underneath it.

The annual space heating SCOP always comes out at the manufacturer's declared figure. The bins decide when the energy is used. They never change how much.

That rule matters because the inputs to the distribution are weather reanalysis and a calibrated physical shape. Any error in them can only move energy between months. It cannot inflate the headline efficiency or the annual running cost.

1. Hours at each temperature, by month. Ten years (2015 to 2024) of hourly ERA5 reanalysis for one MCS design-region station each, via Open-Meteo, binned at 1 °C up to the 16 °C heating threshold.⁴ The same data gives the mean temperature and share of heat for each hour of the clock, which is what splits a time-of-use bill.
2. Heat per month, then per bin. Monthly space heat comes from the project's degree-day profile, unchanged from before. Within a month it is spread across the bins on the EN 14825 load line: hours at that temperature multiplied by the part load the house needs at it.
3. Efficiency against outdoor temperature. A Carnot shape at the project's own weather-compensated flow temperature. The flow falls 12 K between the design condition and zero load, a swing calibrated against monitored UK systems (more below). The curve is held at +12 °C and never extrapolated into the summer.
4. Cycling at low load. Below its minimum modulation a unit cycles on and off. HeatLens applies the EN 14825 degradation, COP × CR / (Cdh × CR + 1 − Cdh), with the regulation's default Cdh of 0.9 and a minimum output of 30% of capacity unless the product record says otherwise.² ⁸
5. Physical ceilings. No bin may exceed 0.6 of Carnot efficiency (the 99th percentile across 199 metered UK systems is 0.59) or 1.6 times the declared SCOP.³ These are backstops that stop a runaway curve reaching a customer report.
6. Pin the year to the declaration. One scale factor for the whole year is solved so that annual heat divided by annual electricity equals the declared SCOP at the design flow temperature. Relative efficiency between bins is untouched.
7. Hot water on its own terms. Hot water is charged at 55 °C whatever the weather (MIS 3005-D 3.5.5), so it gets its own stream: demand from MIS 3005-D and SAP 10.3 Appendix J, cylinder and primary losses, a weekly 60 °C pasteurisation cycle, and an efficiency pinned to 0.9 of the declared W55 figure, calibrated on 183 monitored systems.³ ⁵ ⁶

The house

We use the same reference case as our post on survey input sensitivity: a 1960s semi-detached house of around 110 m² with a filled cavity, 270 mm of loft insulation and a room in the roof, in the Wales design region (Cardiff weather, −3.1 °C design temperature). The HeatLens design heat load is 4.93 kW. The layout is reasoned from records, so treat it as a realistic test case rather than a design for any address.

For this post we pair it with a 6 kW R32 monobloc heat pump whose manufacturer declares an EN 14825 SCOP of 4.47 at 35 °C and 3.26 at 55 °C, giving 3.87 at a 45 °C design flow. Three bedrooms give an MIS 3005-D hot water demand of 180 litres a day with a 210 litre cylinder.

Hot water is 32% of the heat here, above the 15% median across monitored systems. The MIS allowance of 45 litres per person per day is generous for many households, and this house's space heating demand is modest. Both are worth checking on a real project.

What the efficiency curve looks like

The chart below is the heart of the model. The bars are the hours Cardiff spends at each outdoor temperature in a typical year. The line is the space heating COP the model assigns to each bin once the year has been pinned to 3.87.

• Hours per year in bin
• Bin COP

Two details are worth pulling out. The first is how little of the year sits near the design temperature: only about 70 hours a year at or below −1 °C in the ERA5 decade. Sizing is about those hours. Running cost is mostly about the 7 to 12 °C bins, which carry half of the year's space heat.

The second is the downturn at the mild end. An earlier version of this model let efficiency keep climbing into late spring, and gave summer months COPs near 7. No monitored system shows that. Real machines hit a floor on their output, spend mild hours cycling, and carry pumps and controls whose consumption does not shrink with the load. Holding the curve at +12 °C and applying the EN 14825 cycling correction brings the model back into line with measurement.

Checked against monitored homes

We calibrated the weather compensation swing against ten mid-range systems on heatpumpmonitor.org, which publishes metered heat and electricity from UK homes.³ In those homes, the mildest month that still carries real heat (at least 15% of that system's peak month) runs at 1.14 to 1.26 times the system's annual figure, median 1.18. Swings of 8 to 30 K were tested and 12 K gave the closest fit. For our reference house, the mildest month carrying real heat is September at 1.16 times the annual figure, and January sits at 0.92 times it.

Month by month

• Space heating
• Hot water
• Combined (as billed)

Now the electricity. The chart compares the bin model with the usual flat approach, where space heating is divided by 3.87 and hot water by its annual COP of 2.93 in every month. Both add up to the same 3,865 kWh over the year.

• Flat SCOP estimate
• Bin model

In an average Cardiff year the shift is real but moderate: December to February carries 92 kWh (6%) more electricity than the flat model says, about £23 on a flat tariff. If that were the whole story, a flat SCOP would be a tolerable simplification.

It is not the whole story, because averages are exactly what a householder does not experience. Months are made of days, and a cold spell pushes a lot of heat into the bins where the heat pump is least efficient.

The December bill

Suppose December turns cold, averaging around 0 °C for the month, as much of the UK did in December 2010. Heat demand roughly doubles. The question is what that does to the bill, and how well each model would have predicted it.

Typical December, as estimated547 kWh at a combined COP of 3.54

£150

Cold December, flat SCOP943 kWh: more heat, same efficiency

£247

Cold December, bin model1,201 kWh: more heat, lower efficiency

£311

£0£100£200£300

£64 of the cold-month bill is hidden by a flat SCOP.

Against the typical December estimate of £150, the cold month bill is double. About 60% of the increase is extra heat that any honest model should show. The remaining 40% is efficiency, and a flat SCOP hides it completely. That is the part that makes a working system look broken.

This is one of the quieter reasons people complain that heat pumps are expensive to run. The system is doing what physics allows, but the estimate that set their expectations described a machine that does not exist in January. A monthly profile that shows a £167 January next to a £42 July, and explains why, is a much better conversation to have before installation than after the first winter.

Tariffs: when the electricity falls, and at what price

Time-of-use tariffs change the question from "how much electricity" to "how much electricity, when". Until this update HeatLens, like many tools, assumed 40% of space heating electricity would land in the cheap window. There was no source behind that number.

The model now works it out from each region's hour-of-day profile: how much of the month's heat falls in each hour, and the mean temperature at that hour, which sets the COP. Nights are colder, so they take a slightly larger share of the electricity than of the clock.

Share of the clock, 5 hours

20.8%

Bin model, January

22.6%

Bin model, whole year

25.1%

Previous assumption

40.0%

0%10%20%30%40%

Two things stand out. The first is that time-of-use tariffs still win for this house, by £162 to £205 a year, but most of that saving comes from moving hot water overnight and from summer, when the bill is small anyway. In January the Go tariff saves just £6, because the heat pump is running through the expensive day hours at its lowest efficiency.

The second is the size of the old error. On the Go tariff, the 60/40 assumption put the energy cost at £692 a year. The bin model puts it at £791. That is a £99 overstatement of the saving, enough to change the recommendation for a household on the margin, or to make a tariff switch look like a failure a year later.

What would change the answer? Deliberately shifting space heating into the cheap window, with a buffer, a thermal store or a battery. The model gives the numbers needed to size that, which is where it goes next.

Storage is sized for a cold day

Anyone sizing a thermal store or a home battery to move heat pump load into a cheap window needs the energy on a specific day, and usually on a cold one. An annual SCOP says nothing useful about that day. The table uses the bin model's COP at each daily mean temperature, and Cardiff's January hourly profile to split the day.

Flat SCOPBin model

Design day, −3.1 °C

23.7

34.6

Cold day, 0 °C

19.8

26.2

Average January day, 5.8 °C

12.6

13.1

Mild day, 10 °C

7.4

6.2

010203040 kWh

The thermal side is harder still. To carry the whole of the design day's space heating through the 19 expensive hours takes around 92 kWh of stored heat. At a usable 20 K temperature swing, water holds about 0.023 kWh per litre, so that is roughly 4,000 litres. On an average January day it is still around 2,100 litres. A typical 200 to 300 litre buffer holds 5 to 7 kWh at the same swing, which covers a little over an hour of design-day load.

That does not make storage pointless. It means storage is a partial shift, sized against a known cold-day profile, and the bin model is what gives that profile. It also shows why a battery usually does more of the tariff work than a buffer: electricity stored before the heat pump is more compact than heat stored after it, especially on the days when the COP is at its lowest.

Oversizing, cycling and the mild-weather problem

The bin model also makes something visible that a flat SCOP cannot: how many hours the unit spends below its minimum output. An inverter heat pump modulates smoothly down to a floor, typically somewhere between a fifth and a half of its rated output.⁸ ⁹ Below that floor it has to switch off and on to match the load.

Some cycling in mild weather is normal and not harmful in itself.⁹ The problem comes when oversizing pushes the cycling point into colder weather, where the unit is doing a large share of the year's work. More starts mean more wear on compressors, contactors and fans, more time in the inefficient start-up phase, and less stable flow temperatures.

Our input sensitivity post showed that leaving this house's fabric at year-of-build defaults gives a design load of 6.78 kW, about 38% above the evidenced 4.93 kW. The chart shows what oversizing does to cycling, holding the unit's efficiency level at what a correctly sized unit achieves.

Minimum output 30%Minimum output 45%

Matched, 4.9 kWcycles above 13.8 °C

18%

+20%, 5.9 kWcycles above 12.3 °C

27%

+38%, 6.8 kW (defaults)cycles above 11.0 °C

37%

+60%, 7.9 kWcycles above 9.4 °C

57%

+100%, 9.9 kWcycles above 6.5 °C

74%

0%25%50%75%100%

Unit capacity	Oversize	Cycles above	Hours cycling	Heat while cycling	Space SCOP
4.9 kW	none	13.8 °C	2,884 h	18%	3.87
5.9 kW	+20%	12.3 °C	3,593 h	27%	3.85
6.8 kW	+38%	11.0 °C	4,244 h	37%	3.82
7.9 kW	+60%	9.4 °C	5,322 h	57%	3.78
9.9 kW	+100%	6.5 °C	6,050 h	74%	3.70

The pattern is clear. With a 30% floor, a correctly sized unit cycles only above about 14 °C, and 18% of the year's heat is delivered while cycling. At the 38% oversize that fabric defaults would produce, the cycling point drops to 11 °C and more than a third of the heat is delivered in on-off operation. Double the capacity and it is three quarters. With a less flexible 45% floor, a unit twice the size of the load cycles in almost every hour it runs.

On EN 14825's degradation formula alone the efficiency cost looks modest: 1% at 38% oversize, 4 to 9% at double. We think that understates it. The standard's correction covers the energy lost in start-up; it does not model defrost timing, pump overrun, or the wider temperature swings that come with repeated starts, and it says nothing about wear. What the model shows with confidence is exposure: how many hours, at which temperatures, the unit is being asked to do something it was not designed to do smoothly. That is the right trigger for a conversation about sizing, a lower minimum output, or a buffer.

What the method does not claim

A model that redistributes energy needs to be clear about what it knows. These are the current limits.

• The shape is calibrated against measurement. Manufacturers publish COP at −7, +2, +7 and +12 °C under EN 14825, but on the standard's own compensation curve, which drops the flow temperature about 25 K across that range. Used directly, those points overstate mild-month efficiency. HeatLens keeps them for validation and will use them once they can be remapped to each project's own flow temperatures.
• The calibration set is small. Ten systems set the weather compensation swing; 199 set the Carnot ceiling; 183 set the hot water factor. We will keep widening it.
• Daily figures use daily mean temperatures. Real cold days include colder nights and defrost cycles. The design-day figures above are, if anything, optimistic.
• The oversizing analysis holds efficiency fixed. In the app the annual SCOP stays pinned to the declaration. The oversizing figures in this post apply the same efficiency level to a larger unit to show the cycling exposure. They are a separate analysis, and the app's headline figure is unchanged.

Who benefits

Homeowners

• A monthly running cost that shows a realistic January, so the first winter bill matches what you were told.
• Tariff comparisons built on when your heat pump actually uses electricity.
• A clearer view of whether a battery or thermal store will pay its way.

Installers and designers

• Headline SCOP still traceable to the manufacturer's declaration, so reports stay defensible.
• Bin-by-bin heat and electricity for sizing storage against a cold day.
• An early warning when a unit will spend much of the year cycling.

The bottom line

A declared SCOP is a good number. It is independently tested, comparable between products and right for the year. It was never meant to describe January, a cold week, or the five cheap hours of a tariff. Using it that way gets the annual bill right and gets almost every decision that depends on timing wrong: the winter bill, the tariff, the storage and the sizing.

HeatLens now keeps the manufacturer's figure for the year and spreads it across the weather the house will actually see. The annual number stays the same. The months, the days and the decisions get much closer to what a monitored heat pump really does.

See the monthly profile for your own house. Start a project in HeatLens: start from your floor plan and EPC, choose a heat pump and tariff, and see month by month where the running cost falls.

Method and sources

All figures come from the HeatLens annual performance model (September 2026) run on the anonymised 1960s semi-detached reference case: design heat load 4.93 kW by BS EN 12831-1 via the CIBSE Domestic Heating Design Guide with MCS MIS 3005-D rules; Wales design region, −3.1 °C; 45 °C design flow; a 6 kW monobloc with declared EN 14825 SCOP of 4.47 (W35) and 3.26 (W55); three bedrooms, 210 litre cylinder, weekly pasteurisation at 60 °C. Tariff rates are the presets held in HeatLens in September 2026. The cold December is simplified to a uniform 0 °C month. Daily storage figures use the COP at each daily mean temperature and the Wales January hourly profile (22.0% of heat and 22.6% of space heating electricity in the 00:00 to 05:00 window). Oversizing figures divide each bin's load ratio by the oversize factor and hold the efficiency scale of the correctly sized case. Figures are indicative and depend on inputs taken from EPC and floor plan information, which may contain inaccuracies. Designs must be checked and surveyed by a qualified MCS installer.

1. BS EN 14825:2022, Air conditioners, liquid chilling packages and heat pumps: testing and rating at part load conditions and calculation of seasonal performance (bin method, part-load ratio, load line).
2. Commission Regulation (EU) No 813/2013, ecodesign requirements for space heaters, Annex III (reference heating seasons, degradation coefficient Cdh, default 0.9).
3. OpenEnergyMonitor, HeatpumpMonitor.org, public metered heat pump performance data (accessed 22 September 2026; H4 boundary, MID-metered systems).
4. Open-Meteo, Historical Weather API, ERA5 reanalysis from the Copernicus Climate Change Service, hourly 2 m temperature 2015 to 2024, CC BY 4.0.
5. MCS, MIS 3005-D (2025): 45 litres per person per day hot water allowance, 55 °C hot water capability at design conditions.
6. BRE, SAP 10.3, Appendix J (hot water energy content, cold water temperatures, monthly factors) and Tables 2 and 3 (cylinder and primary circuit losses).
7. CIBSE, Domestic Heating Design Guide (2026), and BS EN 12831-1:2017 (design heat load).
8. Viessmann, Vitocal 150-A product literature: output ranges from 2.1 to 4.0 kW up to 3.3 to 14.9 kW, a turndown of 22% to 53% of capacity.
9. Energy Stats UK, Heat pump cycling: normal cycling, short cycling and modulation (minimum outputs by model; mild-weather cycling in context).

If you think any figure here is wrong, we would rather know. Corrections to info@heatlens.co.uk.