The DATA MINING CUP 2015 deals with the topic of coupon generation. What is interesting in this case is the impact of coupons on the shopping basket value as well as the redemption rate of the individual coupons. Coupons have been used as purchase incentives for some time now, not only by mail order companies. This presents interesting challenges, including the question: “Who responds to coupons?”, and then the question: “Who would have made the purchase even without the coupon?”. Only by giving equal consideration to both of these questions can we arrive at profitable couponing. Data mining processes, in particular scoring processes, can be used to answer these questions. Such processes make it possible to create forecasting models which provide answers to both questions.
Using the historical order data from an online shop with accompanying coupon generation, a model should be created that comes up with a prediction for the redeemed coupons and for the shopping basket value for new orders within the shop. The historical data contains both a time stamp of the coupon generation and the orders as well as different product and customer attributes. The information “coupon redeemed yes/no” and the basket value of the order are also known for the historical data.