What is Retail Location Modelling?
‘Customer meeting points’, or stores are substantial investments. Choosing the right store type for the right location can have a big impact on profitability. Location models predict the impact new store openings will have, both new store sales and the cannibalisation effect on the existing store network and online sales.
We can use these models to assess the profitability of candidate new locations. We can also simulate the impact of store closures on overall sales. We can use these models within a data visualisation dashboard. This typically includes interactive maps, showing predicted sales and cannibalisation effects across regions. These insights are hugely valuable for decision-makers trying to assess how to grow their retail presence.
What data is required?
At a minimum we need to know the locations of existing stores and their sales, as well as information about costs. Ideally, we would also have features describing store characteristics - size, range, parking availability etc. And where available time series data providing sales of each store and online by week / month / year and the dates stores were opened / closed.
Beyond this we can gather data on travel times from the Google Maps API and geographic data on population, demographic and economic variables.
How does it work?
The approach depends on the data available and the use-case. In instances where we have aggregated, cross-sectional data only, we can rely on theory (so-called ‘gravity’ models) to simulate store openings and closures. In cases where we have lots of historic examples of store openings and time series data available, we can learn from these. Models then predict the impact of proposed new stores based on these past examples.
Where have we used it?
We’ve used Retail Location Modelling in the retail sector.