← All insights Insight

Why commercial heat pumps underperform after installation

Jonathan Aird · 3 June 2026

A heat pump scheme is signed off on a seasonal efficiency figure. It is commissioned, handed over, and then in most buildings nobody ever checks whether that figure was achieved. When somebody does check, the answer is often disappointing — and the reasons are consistent enough to be worth listing.

We design heat pump systems, and we also monitor them once they are running. That combination is unusual, and it has taught us something uncomfortable: the difference between a scheme that hits its design efficiency and one that misses by a third is rarely the equipment. It is almost always something in how the system was set up, sequenced or scheduled — and it is almost always fixable.

What "underperforming" actually means

A design might promise a seasonal coefficient of performance of 3.2. That means for every unit of electricity the plant consumes, it delivers 3.2 units of heat. If the system actually runs at 2.4, the building is using a third more electricity than the business case assumed, every winter, for the next twenty years.

At current prices that is not a rounding error. UK electricity costs roughly four times as much as gas per kWh, so the running-cost case for a heat pump is already tight. A system that misses its design efficiency does not just underdeliver on carbon — it can turn a marginal financial case into a clearly negative one.

The frustrating part is that this is measurable. Heat delivered and electricity consumed are both meterable quantities. The reason it usually is not measured is that a building management system is designed to control plant, not to prove outcomes. It will tell you a heat pump is running. It will not tell you what it cost to run it.

The six causes we keep finding

Flow temperature set higher than necessary

This is the largest single factor and it is nearly always present. A heat pump's efficiency falls as the water temperature it produces rises. On one scheme we modelled the same plant in the same building at three different flow and return temperatures: at 75/65 °C the seasonal efficiency came out at 2.58, at 65/55 °C it was 3.03, and at 55/45 °C it reached 3.54. That spread was worth about £11,000 a year in running cost.

Systems end up running hot for two reasons. Either the emitters were never assessed at the intended temperature, so the building genuinely cannot be heated any cooler, or — more often — nobody adjusted the weather compensation curve after commissioning and the plant simply defaults to its maximum.

Plant running outside occupied hours

Heating a school at three in the morning is expensive with a gas boiler and worse with a heat pump, because the outside air is at its coldest exactly when the building is empty. We regularly find schedules that were set during commissioning to prove the system worked and never revised, or optimum start settings left at defaults that trigger a full-output warm-up far earlier than the building needs.

Backup boilers and immersion heaters doing the work

Most retrofit schemes retain some form of backup, and quite reasonably. The problem is when the backup runs far more than intended. An electric boiler or immersion heater has a coefficient of performance of one: every unit of electricity gives one unit of heat. If it is covering 15% of the load instead of the 2% the design assumed, it will drag the whole system's measured efficiency down regardless of how well the heat pumps are performing.

On a scheme we modelled recently, the electric backup was designed to run 192 hours a year and deliver under 2% of the annual heat. Whether it actually does is exactly the sort of thing worth watching.

Poor sequencing between heat pumps and existing plant

Where a heat pump has been added alongside retained boilers, the control strategy decides everything. If the boiler is allowed to fire whenever the heat pump cannot meet demand instantly, it will fire constantly, and the heat pump becomes an expensive trim. The heat pump should lead and the boiler should top up, with enough thermal buffer that the heat pump is not being asked to respond to every fluctuation.

Low temperature difference across the circuits

A system designed for a 10 K difference between flow and return but actually running at 4 K is moving twice the water for the same heat. That means more pump energy, more cycling, and a heat pump operating outside the conditions its efficiency figures were quoted at. It usually points to over-pumping, bypasses left open, or balancing that was never completed.

Sensors, meters and controls that are not telling the truth

The least glamorous cause and a surprisingly common one. A flow sensor in the wrong position, a heat meter reading in the wrong units, a valve reporting closed while it is passing. None of it is visible from a plant room walk-round, and all of it corrupts both the control strategy and any attempt to measure performance afterwards.

Why it goes unnoticed for years

Nobody is at fault for most of this. Commissioning happens under programme pressure, often in summer, when a heating system cannot be properly tested. The people who set the system up are gone by the first cold snap. The people operating the building have no way of knowing what good looks like, because the only reference point is a figure in a report they may never have seen.

And the failure is quiet. A system running at 2.4 instead of 3.2 still heats the building. Nobody complains. The cost appears in an electricity bill that has a dozen other reasons to have gone up.

What measuring it actually looks like

The useful number is delivered heat divided by electricity consumed, calculated from metering rather than modelled, over a period long enough to mean something, and compared against the figure the scheme was approved on.

On one site we monitor, a recent week showed a heating coefficient of performance of 3.50 against a design figure of 3.20, hot water at 2.81 against 2.60, and the two combined at 2.98 against 3.00. Two of those are ahead of design and one is marginally behind. That last number matters as much as the first two: a monitoring report that only ever contains good news is not evidence of anything.

Alongside the headline figure, the useful diagnostics are efficiency plotted against outside air temperature, which shows how the system copes as the weather turns, and daily heat delivered against daily electricity used, where the gap between the two is the energy the plant harvested from the air rather than paid for.

The good news

Almost everything on the list above is a settings problem rather than a plant problem. Flow temperatures, schedules, sequencing, compensation curves and balancing can generally be corrected without capital works and often without anyone attending site. The expensive failures — genuinely undersized emitters, plant sized against the old boiler rather than the actual load — are design failures, and those are far cheaper to catch on paper than after installation.

Which is the real argument for measuring. Not to apportion blame after the fact, but because a system that is 20% off design in its first winter can usually be brought back, and one that nobody looks at will stay that way for its entire life.