Webinars
Part 3 – Analytics Rebroadcast: Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement
Data Challenges from Business Disruption – Solutions and Opportunities
Beginning in August 2022, MSI will broadcast each session from our 2022 Analytics Conference weekly on Tuesdays from 12:30-1:00pm ET. Attendees will be able to pose questions to be answered live or within a week of the broadcast. This event is open to corporate members. Please make sure to login to your account in order to register.
Randomized controlled trials (RCTs) have become increasingly popular for measuring advertising effects. However, when RCTs are infeasible, advertisers—or their measurement partners—may rely on observational methods to estimate ad effects. We present the first large-scale exploration of two popular observational methods by comparing their accuracy with 1,673 large-scale RCTs on Facebook. Despite having a wealth of user-level data, neither method performs well, with median relative error rates between 5X and 11X. We conclude that observational methods for estimating ad effectiveness may not work until advertising platforms log highly granular features (e.g., at the auction-bid level) for modeling purposes.
speaker
Brett R. Gordon is the Charles H. Kellstadt Professor of Marketing at Northwestern University's Kellogg School of Management. His research interests include pricing, advertising, and promotions, which he studies using a diverse set of methods from causal inference, machine learning, and empirical industrial organization. He often collaborates with firms to help them measure and design more effective marketing strategies. He currently serves as Co-Editor at Journal of Marketing Research and co-hosts the How I Wrote This podcast, which helps demystify how great marketing papers came to be. Before joining Kellogg, he held faculty positions at Columbia Business School and visiting positions at Chicago Booth and Stanford GSB. He earned his Ph.D. in economics from Carnegie Mellon.