Marketing Mix Modeling (MMM) open-source solutions: A comparative analysis of Robyn, Meridian, and PyMC Marketing libraries.

M. Franco Pérez, X. Puig Oriol

Assessing advertising profitability is critical for data-driven marketing. Marketing Mix Modeling (MMM) quantifies the impact of advertising expenditure on KPIs—typically sales—by accounting for adstock (lag-stocked effect) and saturation (non-linear transformation). These models incorporate temporal components, such as trend and seasonality, alongside external covariates as controls.

With the rise of open-source frameworks, three libraries have gained prominence among practitioners: Robyn (Meta®) [Machine Learning], Meridian (Google®), and PyMC Marketing (PyMC Labs®) [both Bayesian approach]. This work studies the modelling alternatives available within each solutions and details their respective estimation procedures. Finally, we present a comparative performance analysis across various simulated scenarios, offering insights into their behavior, relative strengths, and the persistent statistical challenges for each solution individually and for the MMM domain as a whole.

Keywords: Marketing Mix Modelling (MMM) Sales Attribution Lag-Stock Models Non-linear Saturation effect.

P. A. Cano Chica, V. E. Yáñez Justicia, I. García Garrido, A. J. López Montoya

M. Cousido Rocha, M. G. Pennino, S. Cerviño

P. Gargallo Valero, L. Lample Gracia, J. Miguel Álvarez, M. Salvador Figueras

L. Moreno