Title: Experimental design when A/B tests fail
We're all familiar with A/B testing -- you give some people one ad, some people another, and you look at their behaviors afterwards. But what if you can't control exactly who sees the ads, or you can't directly measure their behavior? This talk will go over some of the real-world issues that break plain A/B testing and will describe some ways to get trustworthy results out of messy environments.
Peter Foley received his PhD in 2013 from the California Institute of Technology, where his research focused on statistical computing, machine learning, and modeling political ideology and behavior. Mr. Foley was previously a Senior Analyst at Strumwasser & Woocher, LLP, and provided data analysis and microtargeting services in the 2012 elections. Lightbox Analytics advises companies on statistical, computational, and behavioral research problems, and Pivotal Targeting provides political experimentation and microtargeting services.