The breakdown circulating online, roof 30%, walls 25%, air renewal 20%, is not backed by any ADEME publication anyone can retrieve. The one the agency publishes today in "Comment isoler sa maison ?" (Clés pour agir series, July 2025, ref. 012579) is computed on its 2025 EPC observatory for a house built before 1974, and it says something else: walls 31%, air renewal and leakage 27%, doors and windows 14%, ground floors 10%, roof 9%, thermal bridges 9%. The roof has gone from first item to last, not because the physics changed, but because the remaining stock has already had its lofts insulated. On site, that average explains an order of priority; it never replaces the element-by-element calculation of NF EN 12831-1, which alone commits you to an output.
The two breakdowns in circulation, and the one you can quote
The two sets of figures side by side
The left-hand column is the one reproduced by renovation websites, the right-hand column is the one in the current ADEME guide. The gap is not cosmetic: the item the first calls the priority is the item the second ranks last.
| Item | Figures in circulation, no retrievable source | ADEME 2025, EPC observatory, house built before 1974 |
|---|---|---|
| Walls | 20 to 25% | 31% |
| Air renewal and leakage | 20 to 25% | 27% |
| Doors and windows | 10 to 15% | 14% |
| Ground floors | 7 to 10% | 10% |
| Roof | 25 to 30% | 9% |
| Thermal bridges | 5 to 10% | 9% |
Why the older version no longer holds
The left-hand column describes a stock that was never treated. Since then, forty years of insulation programmes have targeted the cheapest measure per square metre, the unconverted loft, peaking under the heavily subsidised offers. The result is mechanical: in the homes you survey today, the roof is usually the only item already treated, and the loss has shifted onto the walls and onto the air.
What "typical house" means in the publication
The guide states its scope, and that is what makes it quotable: "average heat losses of a house built before 1974", source 2025 EPC observatory. It is a statistical average over real assessments, not a simulation on a fictitious house. It says nothing about the dwelling you are surveying, and a heat-loss survey run element by element remains the only thing that gives the split for that particular home.
What the average hides: the real spread from one job to the next
The part played by the fabric, the exposure and the climate zone
Two houses of the same floor area and the same year do not share the same breakdown. Compactness, the number of heat-losing elements, orientation and the design outdoor temperature for the location shift the items by several points. A detached house in a cold zone and a mid-terrace in a mild one have neither the same wall share nor the same air-renewal share.
The items that move most between two comparable dwellings
Air renewal is the most widely spread item, because it depends on a system and on a leakage condition, not on a thickness. Thermal bridges come next, because they depend on the nature of the junctions rather than on an area. Losses through the ground also vary sharply with the ground-floor configuration and the exposed perimeter.
What the whole-envelope indicator tells you, and what it does not
A whole-envelope indicator aggregates the elements into a single value and serves to compare buildings with one another. It does not rank work packages. For that you have to go back down to the item, which is what the Ubat envelope loss coefficient does by detailing every element before adding them up.
From the average to the calculation: what you survey on the visit
The three terms NF EN 12831-1 adds together
The calculation does not work in percentages but in watts per kelvin, then in watts at the design outdoor temperature. Three terms are added, and it is their ratio that gives the breakdown for the dwelling.
| Term | What it represents | What you survey to fill it in |
|---|---|---|
| Fabric transmission | Opaque and glazed elements | Area per element, build-up, U-value and its origin |
| Linear thermal bridges | Junctions between elements | Junction type, measured run, psi value retained |
| Air renewal | Ventilation and infiltration | System in place, flow rates, measured or default permeability |
Measure the air rather than assume it
Air is the item where a default assumption costs the most, because it carries 27% of the average and more on already insulated fabric. The air-permeability test replaces the assumption with a defensible figure, and is run at an intermediate stage so corrections happen before the linings go up.
Thermal bridges: by junction type, not by default value
A thermal bridge is counted by linear metre and by type. The six junction types modelled cover most of the common fabric, and that input is what takes the item from an assumed 5% to the 9% recorded in the published average.
From breakdown to work plan: what puts your liability on the line
Prioritise the envelope before committing to an output
The order of the packages is read off the calculation for the dwelling, not off the pie chart. The sequence that holds is always the same: envelope treated, airtightness controlled, then generator sized on post-works heat loss. From that picture, Argile suggests the works to keep and assembles the works plan.
Coordinate the insulation, ventilation and generator interfaces
Treating the envelope shifts the relative weight of the air and changes duct routing. Setting flow rates and airtightness at the same time avoids rework, then the output is placed on post-works demand and checked against the emitter flow temperature. Argile computes that demand room by room to NF EN 12831-1.
What comes back at you in a dispute
Gaps between study and site come from untraced assumptions, not from the percentages announced.
- A default U-value presented as measured, with no mention of its origin.
- A ventilation rate carried over from the study when the system was changed mid-job.
- An output justified by a stock average instead of a sizing note.
