A methodology for data-driven multi-touch channel attribution in online settings

A multitude of channels are utilized by marketers and advertisers to reach customers in the online domain. However, users frequently interact with multiple channels and touchpoints throughout the course of their "customer journey." The assessment of the extent to which each channel contributes to marketing success and the ways in which channels influence each other remains a significant challenge. The prevalent practice of numerous digital advertisers employing relatively straightforward heuristics (e.g., first- or last-click attribution, where the contribution is attributed exclusively to the marketing channel that immediately precedes the conversion or from which the customer journey originates) lacks the requisite flexibility to adapt to evolving consumer journeys in an omnichannel environment. The objective of this paper is to elucidate the shortcomings of rule-based heuristics and stochastic-based attribution methods and to put forth a predictive modeling procedure for individual-level attribution based on open-source machine learning algorithms. Utilizing exported user attribution data from GA4 BigQuery, we propose a reproducible and scalable machine learning workflow for attribution modeling.