Blog › Comfort & Climate

Heating controls that actually save money

The savings are real but come from four specific mechanisms. Judge any control by whether it delivers one of them in your house.

Heating controls that actually save money

Smart thermostats are sold on savings, and the savings are real but frequently smaller than the marketing suggests — because most of the benefit comes from a handful of specific behaviours, and a smart thermostat only helps if it enables one of them.

Here is where heating money actually goes, which controls change it, and what to do if you already have a thermostat and want it to earn its keep.

Where the savings actually come from

Four mechanisms, in rough order of value.

1. Not heating an empty house. The largest single saving for most households, and the most obvious. It requires knowing when the house is empty, which is the part that is hard rather than the part that is clever.

2. Not heating rooms nobody is in. Zoning. Often larger than the first, especially in bigger houses where a lot of space is heated for no one.

3. Running the system at a lower flow temperature. Unglamorous and among the most effective, particularly with a condensing boiler or a heat pump. Explained below because it is the one nobody mentions.

4. Heating slightly less, slightly better. A degree lower on the thermostat is a measurable saving, and it is more tolerable if the house is draught-free, if the heating comes on before you are cold rather than after, and if radiant surfaces are not freezing.

Everything else — geofencing, machine learning, elaborate schedules — is a means to one of these. Judge a control by whether it delivers one of them in your house.

The flow temperature point, which matters most

This is the change most households have never made and it costs nothing.

A condensing boiler is designed to extract extra heat by condensing water vapour out of its exhaust gases, and it can only do so when the water returning from your radiators is cool enough — broadly below about 55 °C. Many boilers are installed with a flow temperature of 75 or 80 °C, which means they never condense and never reach the efficiency printed on the box.

Turning the flow temperature down to around 55 °C typically improves efficiency by several per cent for free. The trade is that radiators emit less heat at lower water temperature, so the house warms more slowly and, in a poorly-sized system, may not reach temperature on the coldest days. The practical approach is to reduce it gradually over a few weeks and see.

For a heat pump the same principle applies far more strongly, because efficiency depends directly on the temperature it has to produce. Every degree of flow temperature costs measurable performance, which is why heat pump systems use larger radiators or underfloor pipework and run continuously at low temperature.

Weather compensation automates this: an outdoor sensor lets the system raise flow temperature only as much as the weather requires, running cooler on mild days. It is standard on good heat pump installations, available on most modern boilers, and frequently left disabled. If your control supports it, enabling it is likely worth more than any scheduling change you will make.

Setback, and why heat pumps invert the advice

The received wisdom — turn the heating down hard when out and overnight — is broadly right for a gas boiler and broadly wrong for a heat pump.

A boiler can produce heat quickly at high temperature, so letting a house cool by several degrees and reheating it works. The building loses less heat while cool, and the reheat is efficient enough that you come out ahead.

A heat pump is most efficient at low flow temperature and struggles to recover a big deficit without raising it. Deep setbacks force exactly the high-temperature operation you are trying to avoid, and can cost more than they save. The right approach is a shallow setback — one or two degrees — and long steady running.

This single distinction explains a great deal of the confusion and disappointment around early heat pump installations, where boiler-era schedules were applied to different hardware.

Presence detection: harder than it looks

Not heating an empty house requires knowing when it is empty, and this is where smart heating most often disappoints.

Geofencing — using phone location — works reasonably for a single occupant with predictable habits and poorly for households where one person works from home, one has irregular hours, and a teenager's phone is out of battery. It also fires late: by the time you cross the boundary, the house has hours of cooling to recover.

Schedules are unfashionable and actually rather good, because most households are more regular than they believe. A schedule that is right 80 per cent of the time, with a manual boost, outperforms a clever system that gets it wrong.

Motion and door sensors tell you about the immediate past reliably, which suits room-level control better than whole-house.

The hybrid that works: a schedule as the baseline, presence used only to cancel heating when the house is unexpectedly empty, and a manual boost that is genuinely easy to trigger. Using presence to suppress rather than to initiate avoids the arrival-lag problem entirely.

Thermostat placement, which quietly ruins systems

A thermostat controls the whole house based on one room, so where it sits determines everything.

Bad places, all common: in a hallway, which is draughty, unheated and near the front door, so it under-reads and the house overheats; above a radiator, which makes it satisfied instantly; in direct sun; on an external wall; near a kitchen, where cooking heat fools it; and in the coldest room, which makes everywhere else too warm.

The best position is an internal wall, about 1.5 m up, in the room you most want to be comfortable — usually the living room — away from draughts, heat sources and sunlight.

Where the wiring makes a poor position unavoidable, a wireless thermostat or a remote temperature sensor solves it, and this is one of the clearest reasons to choose a smart control: you can place the sensor correctly regardless of where the wires go.

What we would be sceptical about

Learning algorithms as a headline feature. They work best in regular households, which are the households a simple schedule also suits. The overlap between “needs learning” and “learning works” is smaller than advertised.

Savings percentages in marketing. They depend entirely on what you were doing before. A household that already turns the heating down when out will save little; one heating an empty house all day will save a lot.

Cloud-only controls. Heating is the function you least want dependent on an internet connection and a company's continued existence. Insist on local schedules that keep running.

Replacing controls before fixing the fabric. Draughtproofing, loft insulation and closing curtains at dusk are cheaper per unit of comfort than any control, and they make lower temperatures tolerable.

An order of operations

  1. Draughtproof and insulate what you cheaply can. This changes what every later step achieves.
  2. Turn the boiler flow temperature down towards 55 °C, gradually, and see whether the house still heats.
  3. Enable weather compensation if your system supports it.
  4. Get the thermostat position right, or add a remote sensor in the right room.
  5. Set an honest schedule that matches your actual week, with an easy boost.
  6. Add presence only to suppress heating when the house is unexpectedly empty.
  7. Then zone, starting with rooms that are heated for nobody.
  8. Measure. Without gas or electricity data you are guessing about all of the above.

Steps two and three are free, take an afternoon, and in our experience deliver more than most people get from the thermostat itself.

Further reading