Lab 6 - Location-Allocation Modeling

This week's assignment was designed to familiarize us with using network analysis to locate best facilities and allocate demand to those facilities. In the first part, we were tasked with working through a tutorial familiarizing us with the concept of location allocation.

In the second part, we were to perform a location allocation analysis with the Minimize Impedance problem to adjust the assignment of market areas serviced by multiple distribution centers for a trucking company. In this scenario, the trucking company has divided the country into several market areas. Our job was to use location allocation analysis to optimize which distribution centers service which market areas. We were provided with a layer showing 22 distribution centers, another layer for customers, a network dataset, and a layer showing unassigned market areas, which is used during the analysis.

At the beginning of the analysis, I set the FacilityType to have a default value of Required. This forces the analysis to use all the distribution centers. I set the impedance to miles and the option for facility to demand, as the trucking company is traveling from the distribution centers to the customers. We are allowing U-turns and using a straight line as the output type. Solving the analysis shows straight lines that connect each demand point to the facility it is allocated to. At this point, several customers are under a market area serviced by one distribution center, but they would be better served by another distribution center. This is the tricky part -- a spatial join needs to be made between the demand points and the unassigned market areas (tricky because the inputs need to be in the correct order; we want a copy of the demand points, not the marketID). I ran the summary statistics tool on the spatial join output, obtaining statistics for the count of the Facility IDs, and using FacilityID and MarketID as case fields, which creates a count of facilities for every unique combination of the case fields. I ran the summary statistics tool again on this output, which determined the facility with the most customers within each market area, I created a table join between the unassigned market areas layer and this latest table and exported that as a new feature so it would become permanent. From here, I could determine the number of customers for each distribution area both before and after the analysis. Next I wanted to map the changes in the market areas. Below is my map, with the original market areas pre-analysis on the left and the new post-analysis market areas on the right. There are very few changes in the western half of the United States, but several in the eastern half. The largest changes are seen in portions of the Midwest, the Ohio River valley region, and the mid-Atlantic states. Many of the changes seemed to make the market areas more compact and in general closer to the distribution center servicing those areas, so the analysis seemed to have the intended effect. The only weakness that I can see is that we used a straight line output from the distribution center to the customer, which is not realistic. The truckers would obviously follow the roads. Overall, however, this analysis is very useful in optimizing the market areas for these 22 distribution centers.