One of the key outcomes of interest in many online experiments is customer churn. In non-contractual settings like food delivery churn is unobserved, and measuring it is difficult. In experiments, this difficulty is compounded by the often relatively short time horizon of experiments, spanning at most a few weeks.
This session will present and discuss one particular solution to this problem: by modelling the purchasing process of customers, it is possible to estimate a feature’s impact on churn without having to increase the duration of the typical experiment.
Speaker bio: Matthias Lux is a Senior Staff Data Scientist at Deliveroo, where he focuses on improving analysis methods for experiments as well as observational data. Most recently, he has developed and implemented a model of consumer behaviour that allows for more personalised targeting of Deliveroo’s consumer campaigns. With more than ten years of experience, Matthias enjoys developing high-quality solutions often based on ideas taken from the academic literature in order to solve measurement problems that provide substantial barriers for the business.
More data science events to interest you 👀