<aside> 📎

A live document of articles, models, books and podcasts talking about Attribution. Created by Barbara Galiza and Timo Dechau. Last updated June 9, 2025.

</aside>

<aside> 🔥

Interested in learning even more about attribution? Enter the waitlist for the Attribution Masterclass Sprint: our free course on all-things-attribution. The next edition is on June 26th.

</aside>

Marketing Mix Modeling

Articles

https://www.021newsletter.com/p/when-to-use-click-attribution-or-mmm

A comparison of click attribution and marketing mix modeling (MMM), covering their strengths, limitations, and best use cases based on data, budget, and goals.

https://www.imf.org/external/pubs/ft/fandd/2011/12/basics.htm

An introduction to econometrics by the IMF, explaining the statistical foundations that underpin marketing mix modeling and how these methods can be applied to analyze relationships between variables.

https://paramark.com/blog/how-do-you-pick-a-target-metric-to-optimize-marketing-towards

A guide on selecting appropriate target metrics for marketing optimization, discussing the tradeoffs between different KPIs and how to align them with business objectives in MMM.

https://paramark.com/blog/how-much-data-do-you-need-to-run-an-mmm

An analysis of data requirements for effective marketing mix modeling, covering minimum timeframes, granularity considerations, and how to work with limited historical data.

https://getrecast.com/google-lightweightmmm/

Technical deep dive by Recast explaining how Google’s LightweightMMM works. Covers priors, seasonality handling, media decay, and geo-level modeling.

Models

https://developers.google.com/meridian

Google's open-source MMM solution that helps marketers measure campaign effectiveness across channels while addressing common modeling challenges like data sparsity and attribution.

https://github.com/google/lightweight_mmm

An open-source Bayesian MMM library from Google, built with JAX and NumPyro. It emphasizes transparency, speed, and flexibility for marketers and analysts looking to run interpretable models without heavy infrastructure.

https://facebookexperimental.github.io/Robyn/