As you know, the Renaissance was the period when the bond broken with Ancient Greek and Roman civilization during the Middle Ages was re-established, marking the rebirth of enlightenment. The word itself means “rebirth.” By analogy, a similar revival is happening for Marketing Mix Modeling (MMM). It is making a strong comeback in the world of marketing measurement.
What is MMM?
MMM can be summarized as a statistics-based approach used since the 1970s to measure the impact of marketing activities on business outcomes such as sales. In essence, all MMM models are regression analyses. Regression is a method used to measure the relationship between multiple variables. For example, a farmer may want to understand the relationship between crop yield and the amount of pesticide, water, and fertilizer used. CMOs have the same need when it comes to marketing.
Why Did MMM Lose Popularity?
While FMCG companies that couldn’t track online footprints and retailers dominated by in-store sales continuously used MMM, most industries shifted to online business models — and measurement methods followed. As a result, deterministic models such as last-click and multi-touch attribution became widespread, while MMM gradually lost relevance.
Let’s briefly explain the difference between deterministic and statistical inference models.
In a last-click attribution model, when a user’s purchase journey is analyzed, it is known with certainty that the user reached the website or app via the relevant channel in the final step — this is determinism. However, the model does not credit earlier touchpoints that influenced the purchase, making it insufficient for understanding where revenue truly comes from.
This led to the emergence of multi-touch attribution models, which assign credit to all channels the user interacted with along the purchase journey. Although these models use algorithms to assign credit, the user’s journey through those channels is still certain — hence, they remain deterministic.
However, these models cannot measure the impact of offline channels such as TV, outdoor, radio, or print, since such media leave no digital footprint to track. Tools like TV Attribution and MMM are therefore statistical inference models, functional where deterministic tracking is impossible. But these models are harder to implement: they require statistical expertise, complex data collection, and carry operational costs.
Why is MMM Gaining Popularity Again?
We’ve explained the difference between deterministic and statistical models, as well as the convenience and perceived certainty of digital deterministic measurement. Yet here’s a bold statement: No one buys a product just because they saw an ad at some point. Assigning sales credit directly to channels is therefore misleading.
Analytics platforms are great for observing how often sales channels appear along the consumer journey, but in recent years, deterministic digital measurement has become insufficient — particularly for CFOs approving marketing budgets.
Growing privacy awareness, stricter regulations, browser cookie policies, consent requirements, and ad-block usage have caused data losses up to 50% for some advertisers. When half the data disappears, deterministic tracking becomes less useful, and the need for statistical models rises. As a result, marketers are rediscovering MMM.
Google Meridian
In the first quarter of 2025, Google launched its MMM service Meridian as an open-source tool. Anyone can now use it for free, and even integrate its open-source code with other products. However, to use the platform effectively, one must possess a combination of statistical, media, and software skills — a rare mix.
It’s worth noting that Google had already offered an open-source MMM tool called Lightweight. Meridian, however, introduces enhanced features and has received positive feedback from early adopters. The model is based on Bayesian theory, which offers a unique approach to uncertainty:
- Frequentist: “Tomorrow’s temperature will be 17°C.”
- Bayesian: “Tomorrow’s temperature has a Gaussian mean of 16°C and a standard deviation of 1.2°C.”
Thus, Meridian’s regression method allows for more informed decisions by quantifying uncertainty. Moreover, Bayesian models perform better with small or incomplete datasets — as data grows, uncertainty narrows, but analysis remains possible even with limited data.
Where Does Medialyzer Fit In?
As a media technology company specializing in measurement, Medialyzer develops data analyses based on statistical models, visualizes results, and helps advertisers use their media channels more efficiently — much like Meridian.
With expertise spanning statistics, media, and software, Medialyzer stands out as one of the few players fully equipped to use Meridian. We recommend adopting Meridian as a complement to our TV Attribution product.
Thanks to the insights provided by MMM, advertisers can gain a clearer understanding of optimal media mix allocation and the effects of non-media marketing activities or macroeconomic factors on consumer behavior.
To build your first model with Meridian, contact us today at sales@medialyzer.com!
