Learning project on public data.
Is Dallas the right place to serve these customers, and would more sites pay?
I mapped 245 ChemCo customers from an MIT case and solved a p-median model for one to four sites, weighted by gallons delivered.
CustomersWith latitude and longitude2451,520,461 gallons in 2004
Dallas todayDelivery cost a yearUS$690K106.1 mi average
Houston directDelivery cost a yearUS$1.87M287.9 mi average
Best 2 sitesDelivery cost a yearUS$475KDallas and Amarillo
What I built
- Plotted every customer on an SVG map from latitude and longitude, coloured by routing zone.
- Solved the p-median model to optimality with PuLP for one to four sites.
- Turned distance into cost with the case figures: US$3.00 a mile, scaled to the 2004 miles.
- Added formula cards for the p-median model, haversine distance and the cost model.
What the numbers show
The best single site lands in Dallas, so the current warehouse is well placed. Shipping direct from Houston would cost US$1.18M more a year in delivery. A second site saves US$215K a year, which is the most it could cost to run.