Capturing Heterogeneity in Engagement on a Social eCommerce Platform

• Activity Number: 240
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Lynd Bacon*+ and Peter Lenk and Danielle Murray and Jean Durall
• Companies: Loma Buena Associates and University of Michigan and Shutterfly Inc. and Shutterfly Inc.
• Keywords: user engagement ; CRM ; Hierarchical Bayes ; marketing ; big data ; social media

Abstract

Many firms need to understand and manage customer engagement in order to be profitable. This is true in particular for firms that use a "freemium" business model where revenues depend on up-selling users of basic, no-cost digital services. And as with marketing in general, quantifying customer heterogeneity in ways that are relevant to decision-making can provide big returns. But being able to do so in today's digital marketing big data environments presents some particular challenges to providing precise and also timely results. We describe a latent variable model of user engagement on a social media platform for use in a distributed data management environment. Engagement parameter estimation is by way of MCMC simulation procedures. Our approach has roots in psychometrics, but is broadly applicable to any application context in which observable indicators of engagement can be identified for modeling purposes, even in "big data" grid or server cluster environments.