RdSAP Is Being Replaced. Here's What Changes About the Numbers.

In our first post we said something specific about where our figures come from: our estimates are built on RdSAP, the government's own methodology for producing Energy Performance Certificates. That was the honest answer, and it's still true today.

It won't be true forever. RdSAP and SAP are being retired in favour of the Home Energy Model, and the reformed EPCs built on it are due in the second half of 2027. If you're planning retrofit work on the strength of modelled numbers (ours or anyone's), it's worth understanding what actually changes when the engine underneath changes, because it's more than a cosmetic rebanding.

This is a piece about methodology rather than policy. We've written separately about the 2030 deadlines. This one is about the thing doing the calculating.

Why replace it at all

SAP was designed for a world of gas boilers and annual averages. Its core calculation is a monthly steady-state method: twelve monthly figures, regional monthly average weather, a fixed efficiency for the heating system, and hot water demand estimated largely from floor area.

For a gas-heated house with no generation on site and no storage, that's a defensible simplification. Gas boilers run at roughly the same efficiency whatever the weather, and a monthly average loses very little.

It falls apart for everything the grid is moving toward. A heat pump's efficiency depends on the temperature difference it's working across, which changes hour by hour. Solar generation depends on when the sun is up relative to when the household uses power. A battery is defined entirely by when it charges and discharges, a concept a monthly average cannot represent at all. Time-of-use tariffs are invisible to a model with no concept of time of day.

So the replacement isn't tidying up. It's a different class of model.

What actually changes

From twelve monthly averages to 17,520 half-hourly timesteps. The Home Energy Model runs a dynamic simulation across every half hour of the year, following the BS EN ISO 52016-1 approach rather than SAP's monthly steady-state method. Weather moves from regional monthly averages to hourly data.

Heat pump efficiency becomes dynamic. SAP applies a fixed efficiency across the heating season. HEM calculates the coefficient of performance at each timestep from the actual source and sink temperatures. For the first time, the model can distinguish a heat pump running into oversized low-temperature emitters from the same unit running into undersized radiators at high flow temperature. Those are very different installations that currently score the same.

Carbon becomes time-varying. The methodology proposes a variable carbon factor reflecting how grid emissions change through the day, instead of a single annual figure. A home that heats when the grid is clean and one that heats when it isn't stop looking identical.

Solar and storage get modelled properly. Rather than an annual yield figure, generation is calculated at each timestep and compared against demand to establish self-consumption. Battery charge and discharge cycles can be represented. Midday surplus charging a battery for evening use is a thing the model can now see.

Hot water gets granular. Instead of a floor-area formula, HEM models individual draw-off events with stratified cylinder modelling, which distinguishes a well-designed system with short insulated pipe runs from a poorly designed one.

The calculation itself gets centralised. Today, multiple software providers each implement the methodology, which introduces variation between engines. HEM is designed to be delivered as a central calculation service, so every assessment runs identical code and methodology updates propagate immediately.

The change that matters most for retrofit planning

There's one proposed change we'd single out for anyone sequencing measures across a portfolio.

Under the proposed methodology, the fabric metric is calculated using a single standard notional heating system in place of whatever is actually installed. Fabric performance is scored on the fabric alone.

Think about what that undoes. Under the current single-score system, installing an efficient new boiler in a poorly insulated house improves its EPC. The rating rises, the house is still cold and expensive to heat, and the fabric problem is now partly hidden behind a good heating score. Anyone who has looked at a stock list and wondered why a solid-wall property was sitting at band D has met this effect.

Separating fabric from heating means a leaky house reads as a leaky house regardless of what's bolted to the wall. For portfolio planning that's a substantial improvement in signal. It's also a warning, because some properties that currently look acceptable will not look acceptable when the fabric is scored on its own terms.

What this means for numbers you hold today

Three things worth being clear-eyed about.

Rankings can move, not just scores. If you're choosing between measures on modelled return, a model that can see time-of-day effects may order those measures differently from one that can't. This matters most where controls, storage, generation, or tariff structure are in play, which is increasingly everywhere.

Some homes will look better without anything physical changing. Heat pumps in particular are expected to score more favourably, partly because the new model represents them more accurately and partly because the carbon accounting changes. That's a methodology difference, not a building difference. If you're comparing a 2026 estimate against a 2028 one, you are comparing two different instruments, and any narrative about "improvement" needs to account for that.

An RdSAP-era band and a HEM-era band are not the same measurement. They will look similar (a letter and a number), and that similarity is the trap. Reading across them without adjustment will produce confident, wrong conclusions, and in a board paper that's an expensive kind of wrong.

What we're doing about it

We'd rather say this plainly than imply we've already solved it.

We are not going to guess at the new methodology and quietly ship numbers based on our guess. The consultation on the EPC methodology closed in March 2026 and the full specification isn't published. When it is, we'll model against it properly, and we'll be explicit about which methodology any given figure came from rather than silently switching engines underneath you.

One aspect we're genuinely pleased about. HEM is modular, which means an assessment can accept better data where better data exists, rather than RdSAP's largely all-or-nothing defaults. Today, if you know more about a property than the standard assumptions allow, that knowledge mostly has nowhere to go. Under a modular model it does. For a platform whose job is turning partial property data into defensible estimates, a methodology that rewards better inputs is the right direction of travel.

What holds up regardless

The reform has already slipped once, from October 2026 to the second half of 2027, and at the time of writing the exact launch date still hasn't been published. So it's worth knowing which decisions are safe to make now.

Fabric is durable under either model. Walls and roofs perform the same whichever engine scores them, and fabric improvements keep paying regardless of methodology, tariffs, or grid intensity. Under the new metrics, fabric is also scored on its own, so fabric work is the least likely of any measure to be revalued by the change.

Property-level knowledge survives the transition. Age, construction, wall type, glazing, heating system, controls: these are facts about buildings, not artefacts of a calculation method. Both models consume them. Time spent getting your stock data right is time that transfers intact; time spent optimising a plan against a specific band threshold may not.

Don't over-fit to a number that's about to be redefined. A plan built around nudging properties over a band boundary is a plan built on the boundary staying where it is. A plan built around genuinely reducing heat demand works under any scoring system anyone cares to apply.

That last point is close to how we think about this generally. A model is a tool for making defensible decisions about real buildings. When the tool changes, the buildings don't, and the teams who understood their stock rather than their scores will have the easier transition.

Sources

Primary: GOV.UK and DESNZ

• The Home Energy Model: Energy Performance Certificates methodology, DESNZ: the wrapper structure, the fabric metric's standard notional heating system, variable carbon factors, and half-hourly smart readiness simulation
• Reforms to the Energy Performance of Buildings regime: partial government response, GOV.UK: the four headline metrics and retention of the legacy rating

Sector

• Elmhurst Energy: SAP vs the Home Energy Model: the move to 30-minute intervals, modular architecture, and centralised calculation delivery
• Elmhurst Energy: Government confirms EPC reform will be postponed to the second half of 2027
• National Housing Federation: Home Energy Model: consultation timing and sector engagement