Marketing mix modelling 4.0: The superiority of agentic, Bayesian optimised marketing mix modelling over traditional approaches

This research paper explores how an agentic, Bayesian optimisation driven marketing mix modelling (MMM) framework offers significant advantages over traditional, static regression-based methods, advancing MMM from its econometric roots to better meet the needs of modern marketers. Traditional MMM often relies on linear regression with predefined transformations and fixed decay rates, leading to sub-optimal parameter estimations and limited predictive power. Our proposed agentic approach leverages Bayesian optimisation to dynamically discover optimal adstock (carryover) and diminishing returns (saturation) parameters for each marketing channel. This data-driven, adaptive parameter tuning, combined with robust feature engineering for seasonality and automated model logging and drift detection, results in more accurate sales attribution, improved return on advertising spend (ROAS) predictions and enhanced strategic budget allocation. This paper details the methodological advancements, presents the underlying mathematical formulations, and discusses the practical implications and benefits for marketing practitioners seeking to maximise their marketing effectiveness. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.