Case Study 4.2.

Within the scope of strategic initiatives, ALFA’s board of directors concluded that it was extremely important to improve the chain’s sales volume. One of the recommended methods for improving sales was to optimize the arrangement of products on the shelves to increase sales. To perform this task, the board decided to employ a market basket analysis and, after generating the association rules, modify the arrangement of products in its stores.

To verify the validity of this approach before applying it in the whole chain, a pilot study was carried out. The sample was 50 stores in which all receipts from a whole month were registered. The algorithm was parameterized in such a way (adequate values for support and confidence) that it generated rules only for pairs of products. The study resulted in a set of rules, which were divided into three groups:

1. Trivial rules (approx. 60%). Self-evident knowledge, usually previously known and used
2. Incomprehensible rules (approx. 30%). Knowledge that is difficult to interpret, not justified in business, or of no use in practice
3. Useful rules (approx. 10%). Vital and previously unknown knowledge that can be directly used for new arrangement of products

On the basis of the useful rules, modifications were introduced in some stores (in which the volume of sales was being monitored) and compared with test stores (in which products were not rearranged). Apart from the new arrangement of goods, salespeople were given special instructions on how, on the basis of the customer market basket, to convince customers to take advantage of upselling of products located near the checkout. After 3 months, the result of this comparison study convinced the board to make a systematic market basket analysis for all its stores and then implement the results across the whole chain.