'Multi-Touch Attribution: Approaches and the Tradeoffs (And Fallacies) Therein' - Tim Wilson / USA

Every marketer wants to accurately measure the impact of their advertising spend. And “digital” was supposed to make that really easy (especially for digital advertising). But, that promise has rarely been realized---it’s becoming increasingly difficult to track users across touchpoints, thanks to privacy regulations (GDPR, CCPA, etc.) and browser updates that block or aggressively expire cookies. In this session, we will review four different approaches to marketing attribution: heuristic modeling (first touch, last touch, linear, time decay, etc.), algorithmic modeling, media mix modeling (MMM), and randomized controlled trials (RCTs). As part of the review, we will venture lightly, but profoundly, into some foundational statistical concepts: we WILL use the terms 'counterfactual' and 'potential outcome,' and probably even 'unobserved heterogeneity!'

##

Filmed at SUPERWEEK 2022 (January 31 - February 4).

MOTION BACKGROUNDS: Amitai Angor, AA VFX

[dawnqian]

So far the most helpful video I watched on this attribution topic! Thanks for sharing

[AryanPatel-wbtp]

if you used a regular regression model for mmm it will be very inaccurate. You need bayesian hierarchal models to deal with adstock and hill saturation .