Free DPE Simulator 2026

How much should you trust this estimate?

It is the question everyone asks and almost no simulator answers. We measured our engine against official certificates, and here is the result — including where it is weaker.

The protocol: our engine against the ADEME oracle

We took 32,744 real certificates published by ADEME, fed our engine the full geometry of each one — walls, roof, floors, glazing, systems — then compared our result to the officially published label. The engine never sees that answer: it runs the 3CL-2021 method on the dwelling characteristics alone, and the comparison only happens afterwards.

Where our engine is least reliable

One area of weakness, which we prefer to document: homes whose certificate was produced at building level then split per flat. On that subset, our exact-label rate drops to 78.9%.

Part of that gap does not come from the calculation but from the measurement itself: the extract we use for this benchmark does not preserve building-level scale, which the calculation needs. When the certificate file is supplied directly, that information is present and the gap narrows.

This point came out of our own calibration work, not a third party. We publish it because a precision figure without its limits is not a measurement.

The six input mistakes we see most often

Since the reliability of your estimate depends mostly on your answers, it is worth knowing where people go wrong. Here are the most frequent confusions, and what they do to the result.

1. 1

   Counting the garage, cellar or conservatory in the floor area

   The expected area excludes the garage, cellar, unconverted loft, balcony and terrace, along with anything under 1.80 m in height. Another classic confusion is entering the gross floor area, before deducting walls, partitions and stairwells. Since 1 July 2024 the DPE works from a "reference area", which does include heated conservatories — unlike the Carrez measurement.

   What it changes: Consumption is expressed per square metre: an over-generous area mechanically lowers the kWh/m² figure and can display a better label than reality. It is one of the most decisive fields in the whole calculation.

2. 2

   Forgetting to describe the windows

   The form only requires one wall before moving on: nothing stops you validating the building envelope without entering a single glazed opening. It is the easiest thing to miss when filling in quickly.

   What it changes: A missing surface is not replaced by an average value: it is counted as losing no heat at all. Yet a heated home always loses heat through its windows. Consumption therefore comes out understated, and the label more flattering than it should be.

3. 3

   Mistaking plasterboard lining for insulation

   A plasterboard sheet fixed against the wall is not an insulant, and a thick stone wall is not "naturally insulated". Two visual cues: internal insulation sounds hollow and makes electrical sockets stand proud; external insulation thickens the façade and deepens the window reveals.

   What it changes: Across 37 real renovations analysed in our input guide, the median wall heat-loss coefficient goes from 1.86 W/m².K before works to 1.02 after. Declaring insulation that is not there therefore erases close to half the losses through the walls.

4. 4

   Confusing purchase date with construction period

   What is expected is the period the building was completed, not when you bought it nor when it was last renovated. If the home has been renovated, keep the original period and describe the works in the later steps. For an extension, use the period of the main structure.

   What it changes: The period drives the default values of the calculation: insulation, air permeability, equipment performance. Getting the era wrong shifts the entire result, including the items you did not fill in yourself.

5. 5

   On collective heating, entering the whole building’s area

   The heated area to enter remains that of your own home, not the area served by the building boiler. The confusion comes from the certificates themselves, where this field describes the entire collective installation: one case in our internal audit shows 49 m² of habitable area against 1,547 m² of declared heated area.

   What it changes: The ratio between consumption and area becomes wrong, and the result unusable. The simulator warns you as soon as the heated area exceeds the habitable area by 30%: better to correct the entry than to override it.

6. 6

   Leaving the district heating network ID blank

   If your home is connected to a district heating network, enter its identifier — four digits followed by a letter, for example 7505C. It appears on the heating bill, can be obtained from the building manager, and the simulator suggests it from your address. Second pitfall: mistaking a collective gas boiler for a district heating network.

   What it changes: Without the identifier, the calculation applies a national average of 0.105 kg CO₂/kWh, whereas real networks range from roughly 0.05 to 0.25 depending on their share of geothermal or biomass energy. The climate label becomes approximate in both directions — and since the final grade takes the worse of the two, the whole result can tip.

Frequently Asked Questions about DPE Simulation