Causal-driven attribution (CDA): estimating channel influence without user-level data

Filippou, Georgios, Quach, Boi Mai, Lenghel, Diana, White, Arthur, Jha, Ashish Kumar
Journal of Marketing Analytics · Springer Nature · 2026-10-03 · DOI: 10.1057/s41270-026-00535-6

Аннотация

Companies advertising across search, social, video, affiliate, and offline media (TV GRPs, radio, and print) face a fundamental challenge: understanding not only which channels drive sales, but also how they influence one another along the path to purchase. As privacy regulations and platform restrictions limit user-level tracking, marketers increasingly rely on aggregated data, where individual customer journeys are no longer observable, making the marketing effective evaluation substantially more difficult. This paper introduces Causal Driven Attribution (CDA), which works entirely from these aggregate counts. CDA integrates temporal causal discovery (PCMCI) with causal effect estimation via a Structural Causal Model to recover directional channel relationships and quantify their contributions to conversions. We apply CDA to reconstruct the consumer journey as a causal graph, revealing direct and indirect pathways. For instance, how video advertising increases branded search, which subsequently contributes to sales. Using large scale synthetic data that replicates real marketing dynamics, CDA achieves an average relative RMSE of 9.50% when the true causal graph is provided and 24.23% when the graph is inferred from the data. Moreover, we also applied CDA to real-world data, where the evaluation yielded promising results, further demonstrating the practical potential of our approach for marketing attribution.

Читать у издателя →

Похожие статьи