[wodenokoto]

The author ends their analysis a little early, which is a shame, because it is a fun but also very interesting research question.

Next step would be to plot the known/suspected warehouses and give an error measurement for the best predicted locations against those more or less known warehouses.

Further steps could be to look into a pattern of how those known/suspected warehouses are placed in relation to stores, and extrapolate that into the missing warehouse locations.