Predict how much the purchase probability could increase as a result of sending an advertising SMS
X5 uses text messages to encourage customers to shop more. Sending texts is costly, so we would like to target a specific group of clients, typically those who would see this as a compelling event. In other words, the goal is to target those customers who wouldn’t have made a purchase without this communication. The goal is to develop an algorithm that predicts whether or not we should send a text message to the client.
This is called uplift modeling problem, see this video for more details (video is in Russian, you can use auto subtitles).
In uplift modeling tasks, clients from the test sample are ranked using descending predicted communication efficiency. The top 30% are selected from such a ranked list as the “most promising” sub-sample. This sub-sample is used to estimate the average added conversion. In a sense, we calculate the average increase in response to the impact on a client.
Prize fund is 400’000 RUR. The winning team will receive 150’000 RUR; second place gets 100’000 RUR; third place receives 50’000 RUR; fourth and fifth places both get 25’000 RUR. There is also a nomination for the best open-sourced solution with a prize of 50’000 RUR. Teams of up to 4 people from all over the world are allowed, regardless of actual physical location. You are entitled to participate in different teams in multiple Retail Hero contests.