When I had my heat pump installed last year, the installers gave what appeared at the time to be odd advice about how to get the most out of it. Testing has shown the odd parts of that advice to be incorrect. I suspect that the odd advice may have come from the premise that COP (Coefficient Of Performance, or its annualised average, SCOP, Seasonally-adjusted Coefficient Of Performance) is the only relevant measure of efficiency. This idea is mistaken, and I suspect that there may be many people using heat pumps in a sub-optimal way because of this.

The odd advice

The advice that I was given that I thought odd at the time was as follows:

• I should not use thermostatic radiator valves in each individual room; and
• I should leave the heating running even when not in the house during the day.

I decided to use my thermostatic radiator valves in any event, and the installers then set my heat pump (a Vaillant Arotherm 7kW) to a target temperature of 30C so that my Honeywell system that controls the thermostatic radiator valves would have control of the heating. I was advised that the target temperature did not affect the amount of energy that the system would consume. The heat curve was set to 1.0.

The installers stated that the basis for the bulleted advice was that the heat pump runs more efficiently when it is running constantly at a lower temperature rather than cycling, that it will run more constantly with a larger circulation volume (i.e. with all radiator valves open), that running it all day will result in less cycling, and heating from cold is less efficient than maintaining an already warm temperature. It was also stated that heating multiple rooms is more efficient because heat from one room can propagate to the next room, allowing a reduction in the flow temperatures because the radiator in the room that it is desired to heat will not need to reach such a high temperature because the heat will be supplemented by the heat from neighbouring rooms.

The hidden untested premises

What was not made explicit, but which is obviously implicit in the above, is the premise that the heat pump is so much more efficient working constantly rather than cycling (etc.) that that efficiency gain outweighs the efficiency loss in heating unoccupied rooms or heating the house at all at times when it is completely unoccupied.

It could easily be the case that the heat pump does work significantly more efficiently, all things being equal, when heating a larger volume of water than a smaller volume of water, and running constantly all day rather than heating only when an occupant is present, but that that all things are not equal, and that the amount of heat energy thrown away heating unoccupied spaces more than outweighs the gain in thermal efficiency of the heat pump system itself. However, one cannot say with any confidence whether this is the case or not without actually measuring this quantitively, and it was plain that the installers had not done this and had no quantitative basis for the implicit quantitative claims made.

I suspect that these claims are made because, if one measures efficiency only by the COP, following the advice will indeed lead to a better metric, and there is no other metric readily available or well known.

However, the true metric for space heating efficiency is the amount of energy needed in any given period of time to heat the occupied living spaces to a comfortable temperature. All things being equal, the higher the COP, the lower that the energy consumed for a given degree of heating will be - but all things are often not equal. 

To take a hypothetical example: suppose that an imagined heat pump will run at a COP of 4.0 in ideal conditions (heating all radiators all day), but at a COP of 2.0 in suboptimal conditions (heating some radiators for part of the day) on a given day. Suppose that the amount of heat energy that the heat pump will produce in those ideal conditions on that day is 100kWh (meaning, with a COP of 4.0, electricity consumption of 25kWh). Suppose, however, that, on that day, heating occupied rooms requires heat energy of only 30kWh. Heating occupied rooms only gives a COP of 2.0, meaning actual electricity consumption of 15kWh. Despite halving the COP, 10kWh has been saved. Thus, the COP alone is an unreliable measure.

My own experiences, setup and data

For reference: I am a one person household in a 1909 built 2 bedroom mid terrace in London. I work full time, but sometimes work from home.

After the heat pump was installed, I read the manual and quickly discovered that the target temperature does affect the energy consumption, so reduced it from 30C to 20C. I also routinely turned the heating system off (i.e., set it to standby) when out of the house, which I later automated using Home Assistant. Through trial and error on cold days, I was able to reduce the heat curve from 1.0 to 0.55 and still maintain a comfortable level of internal heating.

I adjusted the settings for my thermostatic radiator valves. Now, in whatever rooms that I want to heat to the highest temperatures, I always set the target temperatures for the TRVs about 1-2C above the target temperature for the Vaillant system. That means that the valves for these rooms should remain constantly open while the heat pump is running, letting the Vaillant system's flow temperature maintain the heat by running constantly with a low temperature, as is the optimum way for a heat pump to work (all things being equal). For other rooms, the TRVs are set to a lower temperature so that those rooms heat only to the necessary temperature. For example, the study, where I work during the day when I work from home, is heated to a full comfort temperature, but the bathroom, which will receive only occasional short visits during the day, is heated to 18C. The dining room is heated to 18C for most of the day, but increases at breakfast and lunch' time, and I set it manually to increase when I start cooking dinner. At other times, it is maintained at a lower temperature. 

My data show a significant difference in heating energy consumption on days when I am out (and therefore the heating is turned off during the day) and days when I am home (and therefore the heating is running during the day). 

For example, on the 11th of January 2025, the system was running all day, and consumed 17.7kWh of electricity, with outside temperatures ranging between about -1C and +4C. The COP is shown as 2.0. Likewise, on the 14th of January 2025, the system was running all day, and on that occasion consumed 7.5kWh of electricity with outdoor temperatures of 4-9C with a reported COP of 3.0. On the 15th of January 2025, my data shows that the system was not running in the afternoon from about 1500h to 1900h (presumably because I was out); it consumed 4.9kWh with outside temperatures of 7-9C, with a COP of 3.3. On the 16th of January, the system was not running between 1300h and 2200h, and, in outside temperatures of 7-9C, consumed 3.7kWh with a COP of 3.2. 

Moving a little later in the season, on the 4th of March 2025, the system was not running between 1000h and about 2145h; in outside temperatures ranging from about 2-12C (the lower temperatures when the system was running overnight, but with a lower target temperature), the system consumed 3.0kWh with a COP of 2.7. On the 1st of March 2025, by contrast, the system was running all day and consumed 7.7kWh in outside temperatures ranging from 3-11C and with a COP of 2.8. 

For space heating only, in the year 2025, my total electrical consumption to date (the 9th of September) is 675.1kWh, with an average COP of 2.7. 

I have not specifically gathered data about leaving all radiators on to their highest setting as against my normal regime, but may try that this winter to generate some comparison data. 

Conclusion

While COP is a good measure for comparing the efficiency of one heat pump against another, it does not, by itself, effectively measure the overall efficiency of the system, as total efficiency can also be significantly increased by simply generating less heat, even at a cost of having a lower COP. 

While this should not generally affect people's choice of what equipment to install, it should affect people's choice of how to operate that equipment, and suggests that careful setting up and automation is optimum for minimising the electricity consumption while maintaining heat comfort in the home. 

It would be interesting to see others' data to see the extent to which this holds true over different systems, installs and circumstances.