A/B Testing, A Data Science Perspective

Deciding whether or not to launch a new product or feature is a resource management bet for any Internet business. Conducting rigorous online A/B tests flattens the risk. Drawing on her experience at Airbnb, data scientist Lisa Qian offers a practical ten-step guide to designing and executing statistically sound A/B tests.

• Discover best practices for defining test goals and hypotheses
• Learn to identify controls, treatments, key metrics, and data collection needs
• Understand the role of appropriate logging in data collection
• Determine how to frame your tests (size of difference detection, visitor sample size, etc.)
• Master the importance of testing for systematic biases
• Run power tests to determine how much data to collect
• Learn how experimenting on logged out users can introduce bias
• Understand when cannibalization is an issue and how to deal with it
• Review accepted A/B testing tools (Google Analytics, Vanity, Unbounce, among others)

Lisa Qian focuses on search and discovery at Airbnb. She has a PhD in Applied Physics from Stanford University.