Seyed Hanif Mahboobi (Amazon Web Services), Mericcan Usta (formerly at Group M), and Saeed R. Bagheri (Amazon Advertising) were the authors of the paper.
They argued that a classic problem in advertising “is attributing the performance of multi-publisher advertising campaigns to the channels, platforms, sites, creative content and placements involved, each of which typically exposes overlapping target populations”.
However, they added: “Thanks to the increasing availability and matching capability of digital data, cross-device attribution modeling assigns value to each device of an individual on the purchasing funnel.”
The study, entitled Coalition Game Theory in Attribution Modeling: Measuring What Matters at Scale, articulates the results of a three-month-long digital advertising campaign for a national retailer that included display, paid search and programmatic inventories.
More specifically, the campaign served over 400 million impressions to approximately 120 million users. Of these 120 million users, about 137,000 placed an order and were considered converted users for the purposes of the campaign.
And the study used coalition game theory to discuss the relative (and overlapping) impacts of display media, programmatic media and paid search in the retailer’s digital marketing program.
Paid-search inputs, the paper reveals, “had the highest probability of conversion among cases with only one type of input. Adding programmatic inputs to paid search dramatically increased the conversion rate, whereas adding display had a negative impact on conversion rate.”
Moreover, “paid search appeared to provide more effective reach, and overlaps with display were not the best-performing parts of the paid-search audience. Programmatic and display media still exhibited some synergy”.
The authors further observed that “adding display to programmatic media alone led to a higher conversion probability, whereas adding display to any combination with paid search led to a lower conversion probability”.
Sourced from Journal of Advertising Research; additional content by WARC staff