A Time-to-Event Framework for Multi-Touch Attribution

• Activity Number: 165 - SLDS CSpeed 2
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Dinah Shender* and Ali Nasiri Amini and Xinlong Bao and Mert Dikmen and Amy Richardson and Jing Wang
• Companies: Google and Google and Google and Google and n/a and Google
• Keywords: multi-touch attribution; digital advertising; causal inference; poisson process; time-to-event

Abstract

One of the promises of online advertising has been the ability to tie together ad views or clicks with actual outcomes (e.g. purchases, website visits, etc), also known as conversions, in order to give advertisers more insight into the effectiveness of their ads. In particular, advertisers are interested in understanding the relative contributions of the multiple ads a user may see prior to an observed conversion, a problem known as multi-touch attribution (MTA), in order to adjust their budget and bidding decisions accordingly. This presents a challenging problem due to the incomplete (or streaming) nature of the data, as well as the need to allow an ad's effect to change over time. We describe an MTA system, consisting of a model for user conversion behavior and a credit assignment algorithm, that satisfies these requirements. Our model for user conversion behavior treats conversions as occurrences in an inhomogeneous Poisson process, while our attribution algorithm is based on iteratively removing the last ad in the path.