Identical floor area, identical fabric, identical weather. The only thing that changes is how many surfaces face outdoors — and it moves the answer by more than most insulation upgrades do.
The whole-house figure above sizes a heat pump. It cannot size a radiator, and radiators are what actually stop a heat pump working. Each room gets its own design temperature and ventilation rate, its own external walls, and its own emitter requirement at the flow temperature you chose.
| Room | Use | Floor m² | Ext wall m | Window m² | Roof | Floor | ΔT | Loss | W/m² | Rated emitter |
|---|
Heat loss is the sum of everything leaking out through the fabric — ΣU·A·ΔT, each surface's area times how readily it conducts, times how much colder it is outside — plus the heat carried away by air changing over, 0.33·n·V·ΔT. Add them and you have the kW the house needs on the coldest design day. That number sizes the plant, whatever the plant is.
A radiator's rated output assumes water about 50 K hotter than the room. Run it cooler and output falls with roughly the 1.3 power of that difference — not linearly. A boiler at 70 °C gives a radiator about 85% of its rating; a heat pump at 45 °C gives it about 33%.
So the same room needs roughly two and a half times the radiator. Heat pumps in badly performing installations are usually not undersized pumps in a cold house — they are ordinary pumps pushed to high flow temperatures because nobody changed the emitters, which wrecks the efficiency that was the whole point.
A wall only loses heat if there is something colder on the other side. A party wall between two heated homes has your neighbour’s living room behind it at much the same temperature as yours, so across a heating season it carries essentially nothing. Neither does the ceiling under a heated flat, or the floor above one.
This page used to charge every house for four exposed walls, a roof and a ground floor — it costed a mid terrace as though it were standing alone in a field. One 110 m² house came out at 5.8 kW whatever it was attached to, when the honest spread is 3.6 to 6.2 kW. For a mid terrace that is about a quarter too much, which on a heat pump is the difference between a 5 kW unit and a 7 kW one. An oversized heat pump is not merely a bigger bill up front: it cannot turn down far enough to match a mild day, so it short-cycles and runs at a worse efficiency than the one that fitted. The error costs money twice.
Two smaller corrections came with it, both in the same direction — towards the truth rather than towards a bigger number. The ground floor loses heat to the earth, which sits near 10 °C all winter and not to the outdoor air, so EN 12831 corrects it by roughly half; charging it the full outdoor difference was close to doubling the floor term. Against that, junctions — where a wall meets a roof, or a window meets its reveal — conduct more than the flat areas either side, and a fabric-only sum misses them completely. They are added here at 12% of the fabric loss, which is a typical figure for ordinary construction and can be far higher on a badly detailed one.
A heat pump does not have an efficiency; it has an efficiency for a given lift. What it can manage is a fraction — about 40% in a real installation — of the Carnot limit between the air it takes heat from and the water it delivers, and that limit is a ratio that climbs steeply as the flow temperature comes down, not a straight line. An earlier version of this page fitted a line through the middle of the range. It read about right at 45 °C and said 3.7 at 35 °C, where field data on real installations sits nearer 4.0 to 4.8 — so it understated the entire argument for running cool, which is the one thing this panel exists to make.
The emitter drop is the quiet half of the same lever. A radiator responds to the mean of its flow and return, not to the flow — so a system plumbed for a 5 K drop instead of a boiler’s 10 K keeps the water in the radiator 2.5 K hotter on average for the same flow temperature, and gives roughly 7% more output for the cost of a faster pump. It is the cheapest output on the whole list, and the one most often left on the table.
The design heat loss, rounded up to the nearest available unit — not a rule of thumb per square metre. Oversizing costs efficiency because the unit cycles instead of modulating, and it costs money up front. A well-insulated 110 m² house is often 4–6 kW; a solid-walled Victorian one of the same size can be double that.
Rarely all of them. Radiators are frequently already oversized for their rooms, and the requirement is per room, not per house. Work out each room's loss and compare it with what the existing emitter gives at your intended flow temperature — this tool gives you that multiplier.
Twice, in opposite directions. It sets how much output your emitters give, and it sets the heat pump's efficiency — roughly 2.5% less COP per extra °C. Running hot to compensate for small radiators is exactly the trade that makes a heat pump expensive to run.
Because they are built from different information. The whole-house number assumes a plan shape to get its wall area, and one average air change rate. The room sum uses the wall lengths you entered and a ventilation rate per room use. Compare them as W/m², not kW — the two have separate floor areas, so the kW totals are for different houses until those agree. In a well-sealed home the room sum comes out higher, because an average hides the kitchen and the bathroom.
Ventilation. A bathroom is designed for around 3 air changes an hour against 1 for a bedroom, and it is held 4 K warmer. On a modern fabric that makes ventilation three-quarters of its heat loss, so it needs far more per square metre than its size suggests. It is why bathrooms get towel rails that look oversized and still struggle.
No. A real MCS survey measures every room, every surface and the actual construction. This is for understanding the shape of the problem before anyone quotes — and for sanity-checking a quote that looks wrong.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.