We began by examining the core problem: while the DPE (Diagnostic de Performance Énergétique) provides an official energy-efficiency rating, many residents report that their actual electricity consumption does not correspond to this label. Our research therefore focused on determining why these discrepancies emerge and whether they can be meaningfully quantified using available public datasets.
To investigate this, we carried out a detailed exploration of two primary data sources:
Practically, our first task was to test whether the two datasets could be merged. This required several iterations of data cleaning and format normalisation. We quickly discovered that the datasets cannot be linked on a dwelling-by-dwelling basis, as no shared identifiers exist and consumption data is aggregated to protect user privacy. This constraint redirected our approach: instead of trying to match individual households, we focused on comparing average consumption patterns in specific areas with the typical characteristics of the housing stock in the same areas.
Alongside the technical exploration, we conducted interviews and informal conversations with potential users (students, young tenants, and first-time renters). These discussions revealed practical concerns that shaped the project direction. Users expressed difficulty interpreting their electricity bills and understanding what "normal consumption" means. They often relied on DPE labels when renting, but many had experienced bills that felt inconsistent with the promised efficiency. Importantly, when asked what kind of tool would be useful, most participants emphasised the need for a simple interface that answers a straightforward question:
To contextualise our findings, we also reviewed existing public tools such as ADEME's energy simulator and various commercial platforms offering energy-saving advice. While these tools provide valuable information, they rarely incorporate real consumption data or offer area-specific benchmarks. This reaffirmed the relevance of our focus: using open data not simply to display information, but to translate it into practical benchmarks that reflect local realities.
By the end of the research phase, we had a clear understanding of both the opportunities and limitations of the available data. We identified that:
These conclusions guided the direction of the design and prototyping stages, where we began building an interface capable of helping citizens position their electricity consumption relative to what is typical for their area and dwelling category.