Usecase PSV
Challenge
PSV has an online webstore in which a lot of products are sold to its fans. When a fan looked at the website, historical purchases or products that are similar to products often viewed by this particular fan were not taken into account. Team Datacation was asked to create a recommendation system to generate more sales and show the fans products that better matched their interests.
Process
Team Datacation has created a hybrid recommendation system that takes into account personal historical purchase history, personal characteristics, product-product recommendations, and the time element. A fan who always buys a certain away shirt will immediately be recommended the new away shirt. Because the away shirt is often sold with the PSV away shirt, this product will be shown second. Relatively speaking, fans in the 15-20 age group buy the PSV FIFA version most often. The machine learning algorithm will recommend this product to this target audience when the game comes out, but also when it is sold on sale.
Solution
According to theory, the hybrid recommendation system would score better than the current approach, but to see if this is true in practice, PSV is going to conduct A/B tests. By sending out different emails to certain target groups and comparing this with their standard emails, PSV can find out which approach generates more sales. If the hybrid recommendation system proves to work better in practice, the next step is to actually implement it in the webshop and in the mail campaigns.