Marketing Mix Modeling (MMM)
Marketing mix modeling is a statistical method that uses historical time series data – revenue, media spend per channel and external factors such as seasonality – to model how much each channel, online and offline (TV, radio, print, out-of-home), contributes to business results. Unlike user-level tracking, MMM needs no individual cookies or click data; it works on aggregated figures, today mostly with Bayesian models. Google offers a free open-source MMM framework called Meridian, publicly available since late January 2025; an integration into Google Analytics 360 was announced by Google in May 2026 (as of August 2026).
In practice
MMM suits companies with substantial offline or above-the-line budgets (TV, radio, sponsorship, print), where click-based attribution reaches its limits. For Austrian SMEs with a mainly digital media mix, MMM only pays off above a certain budget and data history – ideally two or more years of weekly data – because the models otherwise stay too vague. Google Meridian is freely available as a Python library (github.com/google/meridian), but setting it up and interpreting the results calls for data expertise or outside advice. In practice, MMM is often used alongside attribution data and holdout tests, an approach known as triangulation, to validate budget allocation quarterly rather than daily. One point worth making to clients: MMM does not deliver real-time optimisation, but strategic channel recommendations after the fact.