Is your home a heat trap in summer?

The 5 criteria of the summer comfort indicator

The indicator rests on five criteria defined by the French order of 31 March 2021, annex 9. Two are eliminatory: if either is missing, summer comfort is "insufficient" regardless of the rest. The three secondary criteria then split the "medium" and "good" levels.

• Eliminatory criterion

  External solar protections

  Shutters, blinds or overhangs placed outside the windows facing south, east and west. They stop solar radiation before it crosses the glazing: without them, the home heats up even when well insulated.

• Eliminatory criterion

  Roof insulation

  Insulation of the upper floor, loft or flat roof. The roof is the surface most exposed to the sun; an uninsulated roof transmits heat straight into the living rooms under the eaves.

• Secondary criterion

  Thermal inertia

  The ability of heavy materials (thick walls, concrete slab) to store the coolness of the night and release it during the day. High inertia smooths temperature peaks; a lightweight build overheats quickly.

• Secondary criterion

  Dual-aspect layout

  Windows on at least two façades of different orientation, allowing natural cross-ventilation. Opening up at night creates a refreshing draught that flushes out the accumulated heat.

• Secondary criterion

  Fixed ceiling fans

  Permanently installed ceiling fans. They do not cool the air but speed up its movement, lowering the felt temperature by two to three degrees for a minimal power draw.

Combination rule (annex 9): both eliminatory criteria must be satisfied, then at least two of the three secondary criteria to reach the "good" level — otherwise "medium".

What the regulatory indicator does not look at

We analysed 11.2 million certificates to understand what the summer comfort indicator really says — and what it leaves out. Two findings came out of it.

First finding: the completion gap is smaller than it looks

The indicator has appeared on the DPE since July 2021, and it is filled in 86.4% of the time. That figure deserves an explanation, because you often read that one certificate in three lacks it. That is true if you count every DPE in the database — but a large share of them are dwelling certificates <strong>generated automatically from the building-level DPE</strong>, for which the regulation simply does not provide for a summer comfort assessment (annex 9 § 9.1). We exclude them: 4.2 million certificates, and most of the apparent gap. Once that correction is made, houses and flats turn out to be level — 86.5% against 86.4% — where the raw gap looked enormous.

Second finding: how the verdicts actually break down

Across the 9,663,670 certificates that do fill in the indicator, here is how the three levels break down.

These 40.2% are the indicator exactly as written on the certificates. We also re-applied the annex 9 rule ourselves to the five criteria recorded in each EPC: the share of "insufficient" then rises to 46.5%. The gap is not a matter of perimeter — it is that 16.4% of certificates display a verdict that does not match their own recorded characteristics. The two figures measure different things, and we publish both. Neither is weighted to represent the whole housing stock.

The blind spot: the rule scores the building, not the climate

Annex 9 assesses five characteristics of the home — external solar shading, roof insulation, thermal inertia, cross-ventilation, fixed ceiling fans. None of them depends on where the building stands. An identical house therefore receives the same verdict in Lille and in Menton. This is a deliberate property of the rule, not a hidden flaw: it measures what the building can do, independently of the climate load it faces.

But the effect on how regions rank is striking. We crossed the declared verdict with an internal score that adds local climate and the urban heat island effect. The ranking inverts: cities among the best scored by the rule turn out to be among the most exposed once real heat is taken into account. This is most visible along the Mediterranean coast, but it is not only a southern matter — Paris ranks 2nd among the most exposed municipalities, at 53 out of 100, combining a mediocre building stock with an urban heat island the rule ignores entirely.

National median: 67. These five cities rank among the best-scored by the rule — beyond 390th out of the 426 cities with at least 3,000 certificates our volumes allow us to rank — and among the most exposed once climate is included. We publish the score rather than a rank: all five sit at exactly 58, within a group of 14 municipalities sharing that score. Separating them by rank would mean publishing the order of a sort, not a measured difference.

The reverse is true too

Roubaix has the worst "insufficient" share among the 426 cities our volumes allow us to rank, at 68.2%. Its building stock is genuinely to blame: older homes, little external shading. Once climate is factored in, the city moves back to 114th out of those 426 — so it stays within the most exposed third, but far from the Mediterranean pack. A poor regulatory verdict therefore does not mechanically signal an unliveable summer, and a good verdict guarantees nothing on the French Riviera. That is the whole limitation of an indicator that only looks at the walls.

The mistakes that skew the test

1. 1

   Counting internal shading

   The criterion only recognises protections placed outside the glazing — hinged or roller shutters, external blinds, overhangs, brise-soleil — on windows facing south, east or west, and on roof lights. North-facing windows are out of scope.

   What it changes: Blinds, curtains or films fitted inside do not count: by the time heat reaches them, it has already come through the glass. And this is an eliminating criterion — counting it wrongly is enough to move a verdict from "insufficient" to "average" or "good".

2. 2

   Ticking "cross-ventilated" for a corner flat

   Windows on two façades are not always enough: the rule requires that no single orientation concentrates more than 75% of the glazed area. A corner flat whose openings nearly all face the same way is therefore not cross-ventilated. The test’s question is a simplification — your answer is taken as given, the test does not measure how the glazed areas are distributed.

   What it changes: A secondary criterion counted wrongly can move the result from "average" to "good". Only the certificate’s XML file, which carries the criterion as the assessor declared it, can settle the matter.

3. 3

   Getting the floor wrong in a flat

   The home’s position in the building alone decides the fate of the roof criterion: on the ground floor or a middle floor it is neutralised; on the top floor or under the loft it becomes eliminating again and, absent a declared insulation quality, it is settled by the construction year — before 1975, the roof is deemed uninsulated.

   What it changes: When the certificate already carries the floor, that data prevails and your answer merely pre-fills it. So it is when the certificate is silent about the floor that an inaccurate answer can swing the verdict between "insufficient" and "good" without any other criterion changing.

4. 4

   Confusing a fan with a fixed ceiling fan

   The criterion targets a permanently installed unit, typically a ceiling fan. A pedestal fan, a model set on a piece of furniture or an air cooler do not count. The question does not offer "I don’t know": it reads as "no" until you change it.

   What it changes: Answering "yes" wrongly adds a secondary criterion and can move the level from "average" to "good". Conversely, leaving the question untouched deprives you of the criterion if you really do have a ceiling fan.

5. 5

   Reading a range as a verdict

   When a criterion is left at "I don’t know", the tool does not invent a value: if the answer could change the level, it shows a range, from worst to best case, rather than a verdict; if it cannot change anything, it settles the matter anyway.

   What it changes: Do not take the upper bound: the worst case is the prudent scenario. The Summer Score shown just below does not penalise an unknown criterion — it is therefore optimistic, and the confidence index beneath it drops to signal this. Importing the certificate’s XML file removes the ambiguity and returns a single level.

6. 6

   Mistaking the estimate for the official EPC

   The test result is an estimate, including when you activate the "what if you did these works" levers. Fitting external shutters does not update the existing certificate: a new EPC is needed for the level to change officially.

   What it changes: Putting forward a level in a listing or a negotiation on the strength of the test alone has no legal value. Only an EPC issued by a certified assessor counts.

Frequently asked questions