From Households to Users: Measuring Individual Users Behavior Based on Household-Level Data

• Activity Number: 576
• Type: Invited
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Xiaojing Wang*+ and Shaun Lysen
• Companies: Google Inc. and Google Inc.
• Keywords: measurement ; panel ; user-behavior

Abstract

TV application: A method to measure TV audiences from set-top box data.

To measure TV audience at a large scale, STB (set-top box) logs data may constitute a much cheaper alternative to TV panel data, provided it is possible to infer user-level TV viewing from the household-level STB data. We have developed a method, using panel data, to decompose household level TV viewing into individual estimates. The method features a soft classification approach using signals from user-level demographics and metadata of TV viewing events. To estimate TV campaign reach by aggregating individual event viewing probabilities, an adjustment method is also proposed to account for the dependence among TV viewing events.

Web Application: A method to measuring individual web behavior from household data.

Many online panels require household members to register for each online session creating potential compliance problems. We have developed a household specific model that probabilistically assigns sessions to household members. Once the model is trained, compliance requirements for the household can be eased.