250.000 customers should be assigned to the shops, optimized via distance.
There are different classes of shops (A,B,C,D) with different amounts of customers to assign (A=1000, B=500, C=300, D=100).
In some areas are a lot of shops in a small circle.
My first solution is:
1. Estimation of distance to nearest shop.
2. Order customers with sorting by shops and distance to nearest shop with the lowest amount of potential costumers first.
3. Select correct amount of costumers with nearest shop considering this sorting and the amount in classes.
After that I got 83% correct classification, 83% of the customers are assigned to the nearest shop. In 17% are assigned to other shops (second best, third best, ...). The Problem is to achiev the correct amount of costumers in the classes. It is not possible to become 100% correct assignments ... but I think it is possible to get more then 83%. You can see it in a map.
Unstated, but presumably you have zip code? You will find useful technical and conference topic-related reference information as well as SAS-hosted documentation at the SAS support http://support.sas.com/ website.
In sum there are 250.000, but 17% are not signed to the nearest shop.
The cause is, if there are a lot of shops at the same location and I firstly fill up the shops with a small costumer density around, I get a lot of wrong classifications. I think I need a re-classification in a second step to change some customers between the shops.