Introduction to Uplift Modeling

  • Dr. Juan Orduz from Wolt

  • Estimate causal effects of a treatment (e.g. marketing campaign) to target customers that are likely to respond.

Motivation

  1. How can we optimally select customers treated by marketing incentives?

Issues:

  • Different types of customers
  • Who Come back due to vouchers
  • Never come back regardless
  • Were on a vacation and wouldn't need a vacation.
  • Temporaility paused activties expecting a voucher :D
  • How do we measure?
  • We cannot "send" and "not send" an incentive to same user

What is uplift modeling?

  • https://causalml.readthedocs.io/en/latest/methodology.html
  • https://www.uplift-modeling.com/en/latest/user_guide/index.html
  • CATE estimates