- A consumer goods company redesigned a portion of their supply network. For certain sales outlets, a distinct distribution system had been in place, and the company, to save money, wanted to eliminate that system and place those outlets into the unified system.
- For those outlets, many of the product orders were written by sales employees on behalf of customers. The supply network redesign would result in the eventual elimination of some of these positions, and the company was concerned about the early departure of too many of these employees, such that it would cause disruption to the processing of orders.
- They desired to put a backup system in place, as a fallback, to be able to continue fulfilling demand uninterrupted.
- We developed a custom order generator to provide the case quantity that would be expected, at the item, customer location level, projected out a number of weeks. The heuristics-based algorithm used recent order history and rules to produce the suggested order quantities.
- Since the number of cases ordered on a weekly basis is typically less than four, and often zero, the algorithm had an easy time predicting actual case orders, as verified through a holdout validation exercise. The variances were within one case 93.6% of the time, and exactly correct 63.3% of the time.
- The system was intended to be a “fail-safe” mechanism, with the hope that it would never be used. It was running in the background on a weekly basis during a six month transition period. In fact, there were occasions where it needed to be used – more than had been expected or hoped. The investment in time to develop the system more than paid for itself in terms of better order fulfilment.