How to use conversion-rate (CVR) as an objective in multi-armed bandit experiments

How to Use Conversion-Rate (CVR) as an Objective in Multi-Armed Bandit Experiments

A step-by-step guide with code examples

Intro

Multi-armed bandit (MAB) has become an increasingly important tool for experimentation and has been widely adopted by the industry giants such as Google, Meta, Netflix, LinkedIn, etc. to conduct efficient experiments. However, widely-used MAB test designs require the objective of interest to provide instantaneous feedback in order to update the assignment probability to each variant. …

a/b testing conversion data science rate reinforcement learning thoughts-and-theory