Changes in consumer behavior have led retailers to pursue higher levels of customer proximity, by increasing the number of stores while focusing on convenience formats. This shift in focus requires heightened agility when analyzing possible new store locations.
This project aimed at improving a retailer’s accuracy when predicting the sales potential of given store locations, while also increase the speed at which these forecasts are generated. The goal was to find the highest value locations for the retailer’s future stores.
We started off by estimating the total market size by combining different data sources, such as external reports, demographic stats and raw datasets (e.g., using Airbnb’s listings to plot the impact of tourism across key urban areas). This allowed us to have a remarkable level of geographic detail without foregoing precision.
Afterwards, we built an interactive app that makes it possible for the user to study the performance of future stores, given specific store characteristics such as exact location, store size and store brand.
This tool has a built-in machine learning model that forecasts store sales by looking at the competitive landscape and estimating the store’s market share within its area of influence
The developed solution enabled the client to increase sales’ forecast accuracy by 15% and to decrease the time needed to analyze one new store location from 5-10 days to 1-2 minutes.
Since a solid forecast could be delivered much faster, the focus of the team shifted to promising locations only. Multiple validation scenarios were tested to prove the robustness of the solution, which allowed it to become a central piece in this retail client’s expansion process.
By: Paulo Sousa , Filipe Figueiroa
