A/B testing is widely recognized as the most reliable method for establishing causal relationships between interventions and outcomes. However, this method requires a randomly assigned control group, which may not always be possible in practice. Additionally, A/B testing may not be optimal when trying to measure the impact of interventions that have already occurred but were not performed according to A/B testing requirements.
In this talk Matheus Torquato (Senior Data Scientist at Jaguar Land Rover) will explore how to measure the impact of an intervention when A/B testing is not possible with Synthetic Control, a methodology that allows the estimation of causal effect of a policy or program by constructing a “synthetic” version of a treated unit. What has been considered as “arguably the most important innovation in the policy evaluation literature in the last 15 years” [Journal of Economic Perspectives, 2017], can help you answer important questions such as the impact of a new policy, the effect of a marketing campaign on sales, and the value of a training program on employee productivity.
In this talk, Matheus will introduce Synthetic Control and showcase its power through a case study of an intervention implemented at Jaguar Land Rover. This method was crucial in persuading other parts of the business about the effectiveness of our new recommendation system for ordering 60% of Jaguar Land Rover stock (150k vehicles).
Whether you are a data scientist working in academia, industry, or government, or simply interested in the latest developments in causal inference, come along to see how this paradigm shift method for evaluating and measuring the effectiveness of interventions can help you persuade with data.
Technical level: Introductory level/students (some technical knowledge needed)
Supported by
