The Certificate and the City
Exhibit 02, from our run of 2026-07-06. The exhibit above carries every number, drawn from the published data.
A certificate records how one machine performed in one laboratory test. The conditions are fixed and the result is repeatable, which is the point of it. What it cannot tell you is how the same machine will behave in a particular home, because a home is not a fixed condition: the weather moves, buildings get extended and re-glazed, and plenty of radiators were sized for a boiler that left years ago.
So we asked a plain question. What does one certified machine do across a whole city of homes, over a whole year?
What we did
We took one certified 8.5 kW air source heat pump and worked it through 24,931 homes across Greater London, each home reconstructed from its public EPC record. Every home ran a full year, hour by hour, through the government's Home Energy Model, against a standard London weather year, at a 55 °C design flow.
Modelled is the operative word. No meter was read and no home was visited. These figures come from the public record and the reference method alone, and the measure is seasonal efficiency in the home: the space heating a home received over the year, divided by the electricity the heat pump used to provide it.
What came out
The certificate's headline figure is 5.0. That is the SCOP at 35 °C flow, flow being the temperature of the water the heat pump sends to the radiators. The hotter that water has to be, the harder the machine works, and 35 °C is the kindest case a test rig can be given.
The median home delivers 3.6.
Nine homes in ten land below the headline figure. A tenth of homes sit below 3.24 and a tenth clear 4.92, so the same machine with the same certificate produces a citywide spread, and the spread is decided by the homes: their fabric, their emitters, their hours.
There is a second certified figure in the same public record, taken at 55 °C flow, which is the water most existing radiator circuits actually need: 3.6. Our run takes the machine's certified performance at that flow as its anchor, so the fair reading is this: the certificate already tells you what the machine does at the temperature your radiators need. What it cannot tell you is which homes will ask for that temperature, for how many hours, through what fabric. That is where the spread comes from, and the spread is the finding.
The reading error is buying on the headline figure when the relevant one sits two lines below it.
If you make heat pumps
The product performs to its certificate throughout, so none of this reflects on the machine, nor on the people who install it. What varies is everything around them: fabric, emitters, demand, hours of use.
The problem it leaves you with is routing. You answer for a base of homes, and one figure for the whole range cannot tell you which of them are ready now, which need evidence first, and which are not yet worth a visit. Sorting that takes a number for each home, before anyone is sent to the door.
If you run housing
A budget built on the headline figure buys 5.0 on paper and 3.6 in the median home. The difference is a promise already made to a board, a funder and a resident before any evidence existed.
The spread matters more than the median. In the same city, with the same measure and the same machine, one street spends well and the next does not. What makes a budget defensible, when the funder eventually asks what it bought, is evidence for each home before the money moves.
About the method
Every number above is modelled: a full year at hourly steps, a London test reference year for weather, a 55 °C design flow. Monitoring data, field trials and installed systems sit outside this study, deliberately: the question was what the public record already implies, before anyone visits.
The homes are reconstructions from public EPC records, and those records age. Any single reconstructed home may be wrong in ways its record cannot show, which is why the claim here is the population rather than any one address.
The product is anonymised deliberately. It performs to its own certificate, and naming it would turn a finding about housing into a story about a brand.
The certificate is doing its job. A rig test is repeatable and comparable, and it only goes wrong when a single point is read as a prediction for a particular home. A certificate is one point; a city is a distribution. The record HeatAssure keeps for a home is the other half: the home's own number, held against what it was promised.
Where this leaves you
The living version of this kind of reading is in our views, where you can touch the numbers yourself. If you answer for a base of heat pumps, start at the OEM door. If you answer for homes and the budgets that improve them, start at Housing & Place. The question in both cases is where your homes land, not what the headline says.
Source notes
- Exhibit 02, generated 2026-07-06: one certified 8.5 kW air source heat pump worked through 24,931 Greater London homes reconstructed from public EPC records, full year at hourly resolution, London CIBSE test reference year, 55 °C design flow temperature
- Certificate figures from the public record for the product: SCOP 5.0 at 35 °C flow, 3.6 at 55 °C flow; the run anchors on the certified 55 °C performance
- Method: the government's Home Energy Model; metric is annual space heating delivered divided by heat pump space heating electricity input