The heating curve is the most decisive setting in a heating installation. It is also the one nobody ever revisits — and the only one whose correction costs nothing.
Every autumn, thousands of buildings restart their heating with their commissioning settings. Sometimes those of twenty years ago.
The 2006 setting that still heats in 2026
The heating curve establishes the relationship between outdoor temperature and heating flow temperature: the colder it is outside, the hotter the water leaves. It is set through two parameters, the slope and the parallel shift, on top of the room setpoint.
At commissioning it is set high. That is rational: the installation has not yet run under real conditions, the building is sometimes still damp, and the one thing an integrator cannot afford is a cold room at handover. A generous curve settles that risk in a minute.
The problem is not that initial setting. The problem is that nothing in the life of a building provides a moment to come back to it. Windows are replaced, insulation is upgraded, offices become meeting rooms, a floor empties out. Demand falls by 20 or 30%. The curve does not move.
How to spot it without instruments
Four signs are enough, and all of them can be observed in a single visit.
- Thermostatic valves are mostly throttled down. If occupants have to restrict emission everywhere, the emission is oversized by the setting, not by the radiator.
- It is too warm in mid-season. Excess in mid-season points to the parallel shift; excess in cold weather points to the slope. The two are corrected separately, and confusing them is the most common tuning mistake.
- Somebody opens a window in winter. It is the most reliable signal there is, and the most expensive.
- The return temperature stays high. On a condensing boiler, a return above roughly 55 °C suppresses condensation: a high-efficiency generator is being run as an ordinary one. On a heat pump, every excess kelvin of flow temperature costs around 2 to 2.5% of coefficient of performance — over a season, that shows up on the bill.
The mistake not to make
A high curve is rarely a calculation error. It is almost always a compensation: one poorly served room complained, and instead of dealing with that room, the curve was raised for the whole building. The setting solved a complaint by degrading everything else.
Practical consequence: do not touch the curve before hydronic balancing and terminal settings have been dealt with. If the critical room is not sorted out, lowering the curve will produce a complaint within three days, somebody will put the setting back up, and the exercise will be filed as a failure.
Balancing first, curve second. A high curve is almost always the compensation for a single poorly served room; until that room is dealt with, any reduction is back to its original value the following week.
The method, step by step
- Record the present state. Slope, parallel shift, room setpoint, night setback, separate curves per circuit if there are several. Write it down: this is the fallback point if the campaign goes wrong.
- Choose a reference room that is representative, rather exposed, with its thermostatic valves fully open for the whole campaign. Put a room temperature logger in it. Without a fixed reference point you are not tuning, you are guessing.
- Log two weeks of outdoor, flow, return and reference-room temperature before touching anything. That record is what will prove the gain, and it is also what reveals a generator that short-cycles.
- Lower in small steps. Two to three kelvin of flow temperature at the design point, that is one notch of slope, per one- to two-week stage. A heavy building takes 48 to 72 hours to respond: judging faster means judging the thermal inertia, not the setting.
- Correct slope and parallel shift separately. Too warm only in mild weather: lower the parallel shift. Too warm only in cold weather: lower the slope. Too warm throughout: lower both, starting with the parallel shift.
- Watch three indicators at every stage: the reference room temperature, the morning recovery time after the night setback, and the number of generator starts. The first one to degrade sets the limit.
- Stop one notch above the breaking point. The right curve is the one where, during a cold spell, the reference room still reaches its setpoint with its valves fully open. No higher.
The safeguards
Three limits are non-negotiable, and they must be written down before starting.
| Safeguard | Rule | Why |
|---|---|---|
| Domestic hot water | Leave it alone | It has its own temperatures and its own hygiene constraints; the campaign concerns heating circuits only. |
| Frost protection and minimum flows | Remain active at all times | Lowering a curve must never take a generator below its minimum flow or disable a protection. |
| Underfloor heating | A few kelvin of margin, longer stages | It already runs at low temperature and its inertia is high. On mixed installations, each circuit is treated with its own curve. |
What it earns, and what it does not
One has to be honest about the order of magnitude. Lowering a curve does not transform a building: it recovers what a setting gone wrong was dissipating. On an overheated radiator installation, the typical margin is 5 to 10 K of flow temperature at the design point, and the gain usually sits in the low single-digit percentage of heating consumption — more on a heat pump, where the reduction acts directly on the COP, and more still on a condensing boiler that starts condensing again.
What makes the exercise worthwhile is not the percentage. It is the ratio: no order, no shutdown, a few engineering hours and a data logger. There are not many energy efficiency measures one can say that about.
And there is a useful side effect. Within three weeks, a curve-lowering campaign brings out everything the installation was hiding: the room that was never balanced, the valve stuck at its end stop, the outdoor sensor mounted in full sun, the circuit whose served area nobody can identify any more. That is often where the real potential lies.
wall-i is an independent engineering office for BMS and building automation, based in Valais, Switzerland. Our specifications are open to every brand and our deliverables follow the SIA standards and the KBOB naming convention. A subject you would like us to cover: hello@wall-i.ch