The EPC is lying to you (a little)

Two houses on the same street. Same age, same walls, same loft. One holds an Energy Performance Certificate that says band C. The other says band D. A buyer skims both listings, feels a small nudge of reassurance towards the C, and never learns that the two certificates are separated by a single point on a 100-point scale — a point that, on a different day, with a different assessor, could just as easily have gone the other way.

The EPC is not a fraud. It's one of the more useful free documents attached to a British home, and we lean on it constantly. But it's a model, not a measurement, and the single letter everyone actually reads hides how soft the number underneath it really is. We analysed all 22.5 million certificates in the national register to show you exactly where the letter tells the truth and where it bluffs.

What the certificate actually measures

An EPC does not measure the energy your house uses. Nobody reads your meter. An assessor spends 30–60 minutes recording the fabric of the building — wall construction, glazing, loft depth, boiler type, floor area — and feeds it into a standardised model (RdSAP) that estimates the cost to heat and light a notional occupancy of that building. The output is a number from 1 to 100. That number is then chopped into letter bands:

Two things follow immediately from that design, and both get lost the moment a listing collapses everything into "EPC: C".

First, it's a cost model with fixed assumptions — standard heating patterns, standard weather, standard occupancy. Your actual bills depend on how you live, and can diverge wildly. Second, and more importantly for anyone comparing two homes: the bands are arbitrary lines drawn across a continuous score. A house can sit one point below a boundary and one point above it and be, in every physical respect, the same house.

The one-point cliff

Here's where it stops being theoretical. The line between band C and band D falls between a score of 68 and 69. We counted how many homes sit right on that fault line:

Nearly 800,000 homes score exactly 68 — the very top of D — and another 946,000 score exactly 69, the very bottom of C. Those two groups are, for all practical purposes, the same house. One point of modelled difference — a slightly more generous assumption about loft insulation, a boiler logged as marginally more efficient — is the entire distance between "D" and "C" on 1.7 million homes.

Widen the lens a little and the crowding is startling. More than a third of every certificate in the country — around 8 million homes — sits within five points of the C/D boundary. And it matters commercially, because C and D are where Britain lives: those two bands alone account for 69.5% of all certificates. The single most consequential line in the whole system is the one drawn straight through the middle of the market, exactly where the homes are most tightly bunched.

If you're weighing up a property and one listing says C while another says D, the honest interpretation is usually not "one is meaningfully more efficient." It's "these two are a rounding error apart, and someone got lucky with a threshold." Read the actual score, not the letter.

The same house, a different letter

Bands aren't just close together — they move. When a home is assessed more than once, the band frequently changes, and not because anyone touched the building.

We took a 1% sample of properties that hold two or more certificates. Among them, 58.9% carry more than one band across their certificates, with a median swing of 7 points between a property's lowest and highest score — and a mean swing of nearly 11. Seven points is half the width of an entire band. A model whose repeat measurements of the same house routinely wander by half a band is telling you something important about its precision.

Some of that movement is real: a home that adds insulation or swaps a boiler genuinely improves, and re-certifies higher. That's the system working. But a swing this large, this often, on unchanged buildings, is the assessor-to-assessor, model-version-to-model-version noise that the confident single letter completely conceals. The band is a snapshot of one afternoon's inputs, not a fixed property of the house.

The "potential" fiction

There's a second number on every certificate — the potential rating — and it's the most quietly optimistic figure in British property. Across all 22.5 million certificates, the average current score is 65 (a mid-D). The average potential score is 79.6 — a high C, nearly a B.

71.9% of all certificates — over 16 million homes — carry a potential band better than their current one. The certificate is, in effect, forever telling homeowners "you could be a C" on the assumption that they'll install every recommended measure at once: solar, a heat pump, wall insulation, the lot. It's not a lie exactly, but it's a sales projection dressed as data, and it's why "improvable to C" in a listing means almost nothing on its own.

So where is the EPC reliable?

We're the last people to tell you to ignore it. Used properly, the certificate is genuinely valuable:

• At the extremes, the letter is trustworthy. An F or a G is a real warning — solid walls, no insulation, electric heating — and an A/B is a real reassurance. It's the enormous C/D middle where the letter stops discriminating.
• The underlying score beats the band. A 79 and a 69 are both "C", but they are not the same house. Always read the number.
• The recommendations are the useful part. The itemised measures, their indicative costs and the fuel breakdown tell you far more about a home's running costs than the headline grade.
• Trend over precision. One certificate is a noisy single reading. The direction across re-assessments, and how it squares with the property's age and construction, is the signal.

The EPC is a screening tool, not a verdict. It's brilliant for ruling out the genuinely bad and flagging the genuinely good, and close to useless for splitting hairs between two mid-market homes — which is exactly the comparison buyers most often try to make with it.

That's the whole reason we build Ask the House: a single letter, modelled on one afternoon and rounded to the nearest threshold, was never going to answer "is this a good house to buy?" The certificate is one input. The flood zone, the conservation-area constraints, the planning history and the price it last sold for are the rest of the answer — and none of them fit on a coloured sticker.

Methodology

Source and scope. The England and Wales EPC register: 22,518,549 certificates, analysed on 18 July 2026. Headline figures are calculated directly from the published records rather than estimated.

• Band distribution: C = 7,430,726; D = 8,215,469; together they make up 69.5% of the 22,518,457 certificates with a valid A–G band. Ninety-one records carrying a stray "I" and one blank are excluded.
• The C/D cliff: scores 64–73 account for 8,017,997 certificates (35.6% of the register). The table above reports the count at each score around the boundary.
• Re-assessment variance: a deterministic 1% sample of properties with at least two certificates (n=48,898). Of those, 58.9% show more than one band; the median spread between lowest and highest score is 7 points and the mean is 10.7. Some movement reflects genuine improvements, so this should not all be read as model error.
• Current vs potential: the average current score is 65.0 and the average potential score is 79.6; 16,189,146 of 22,518,439 certificates (71.9%) carry a better potential band than current.

What the records can't tell us: whether an individual assessor got a specific home wrong. EPC records do not carry metered consumption, so the model cannot be validated against real bills at national scale.

---

Want the rest of the picture on a specific home — not just the letter? Ask the House pulls the EPC, flood risk, planning history and sold price into one grounded answer. For the full guide to what the energy record shows and how to read yours, start with home energy ratings, honestly. Or brush up on the vocabulary: EPC band, EPC certificate, and the wider property glossary and buyer questions.