This project aimed to create service zones for Turf Badger’s Stevens Point branch that satisfied three principles common to territory design problems: balancedness, contiguity, and compactness (Moreno et al., 2020). Achieving these principles required utilization of an optimization data analytics technique in heuristic modeling. The final heuristic combined a Genetic Algorithm with a Manhattan-Distance K-Means Clustering approach. 

Through collaboration with management, it was determined that six zones, each with a minimum income of $20,000 USD and as close to the average available amount of $24,999.17  USD, was necessary to achieve balancedness. As for contiguity and compactness, non-overlapping, tight cluster groups for each zone were desired. Overall, the clustering component of the heuristic first implemented the contiguity principle. Then, the cluster centers were adjusted with the Genetic Algorithm to minimize an objective fitness function for standard deviation in income and an additive radial compactness penalty-like term.  

The solution was developed through grid-search of potential mutation, crossover, elitism, and parent-portion Genetic Algorithm parameters. Ultimately, the optimal model produced a standard deviation in income of only $51.47 with a compactness value of 489.23. All zones met the principal objectives and at most income deviated from the mean by only 0.17%. The solution was implemented into Turf Badger’s FieldRoutes scheduling software for use by the company going forward.