In this instance our client came to us with a plan for a major upgrade for their worldwide Mystery Shopping programme.
As part of the upgrade the client wanted to simplify the way requests to suppliers were handled,
The client makes over 5000 individual shopping requests every quarter, so it was important to them that as much of the process as possible could be generated automatically.
By grouping suppliers being tested together in small sets and giving each supplier in the set the same order, instead of making individual requests to each supplier, we were able to streamline the process, and achieve greater consistency for the client.
Part of our challenge was to identify suppliers to be grouped together, as competitors, and in designing the system to provide a tender description which all the suppliers within the group could reasonably meet.
By using a hybrid approach of a Constraint solver paired with simulated annealing, we found a mechanism which allowed our client to control which scoring factor to adjust over time which gave them the desired fine control over the outputs generated.