Analysis of a retail branch network: a problem of catchment areas

This paper outlines a spatial distribution problem for a national building society and a solution using a genetic algorithm. The mortgage market is highly sensitive to local conditions, the same level of business in cash terms may be very good in the North West of the country but very poor in the South East. When allocating sales targets for managers it is vital to take account of the broad regional differences and more local effects that relate to competition, supply and demand. However, before detailed spatial analysis can take place the catchment areas of the branches must be defined. The genetic algorithm defines catchment areas that minimise the differences between the values of mortgages held by branches in the network and the level of business covered by their catchment areas. The GA adapts a simple gravity model to allocate recorded business to the most likely branches. This is a complex source and allocation problem, similar to multiple facility location, that has been solved using a GA to search through the vast number of potential configurations.