Urban Space
Brazilian real estate data is fragmented and unreliable: every portal has its own slice, and listed prices carry too much noise to support an investment decision.
The hypothesis
Price per square meter can be normalized by the socioeconomic structure of the surroundings. Using public indicators as a baseline, the model builds a price/m² multiplier that captures urban inequality patterns and shows when a property is expensive or cheap for its own address, not for the city average.
- Stack: Python, scikit-learn, BigQuery and pandas over public socioeconomic data;
- Approach: normalization by location instead of direct listing-to-listing comparison;
- Status: internal R&D. It is the methodological base of our real estate products, such as AgregaImĂłvel.
Why we show this
Not every project becomes a product. This is the kind of exploration we run before accepting a data thesis as true.