23 Jun | Zoom at 16:00 CEST / 10:00 EDT
How The Washington Post Uses Reinforcement Learning for Paywall Decisioning
Speaker: Janith Weerasinghe | Senior Data Scientist at the Washington Post | Ph.D. in Computer Science, NYU Tandon | Research Interests: Reinforcement Learning, NLP, and Privacy
Moderator: Dr. Ana Moya | Lead, WAN-IFRA Data Science Expert Group
About the webinar:
Most digital publishers rely on paywalls to convert readers into subscribers, balancing engagement and revenue. In this session, I will explain how The Washington Post transitioned from a rule-based metering system to a reinforcement learning approach for paywall decisioning.
It will share why this problem is naturally modeled as a reinforcement learning task, how a randomized control trial to collect training data was used, and how an offline RL model was trained to make these decisions. It will also cover how models offline were evaluated and deployed them safely in a revenue-critical environment without online exploration.