Lingering privacy challenges and ever-improving cloud and artificial intelligence technology are driving a marketing model renaissance.
Marketing mix modeling (MMM), launched in 1949, fell out of favor in the early 2000s when digital advertising took off. The data-driven technique had long helped marketers understand how variables such as advertising, promotion, and prices impact revenue.
Yet compared to tracking cookies and last-touch attribution models, MMM seemed complex and expensive.
Renaissance
In 2025, however, MMM has enjoyed renewed attention.
Meta and Google have released free, open-source MMM tools in the past couple of years — Google’s Meridian on January 29, 2025, and Meta’s Robyn in 2023.
Why does MMM interest two of the largest digital advertising platforms? I see three probable factors: tracking cookies, AI, and cloud computing.
Google launched Meridian, an open-source marketing-mix model, last month.
Cookie-less Advertising
Controversies surrounding tracking cookies are the first driver. Cookies are a foundational and useful technology. A first-party cookie on a browser keeps users logged into a website and retains their preferences.
However, third-party tracking cookies that catalog an individual’s behavior across web properties are a privacy pariah. Laws such as Europe’s General Data Protection Regulation and the California Consumer Privacy Act limit such cookies, and many browser companies have stopped supporting them entirely.
The potential of cookie-less ad targeting makes MMM attractive to large-scale advertisers and platforms.
Advertising performance. Third-party cookies, despite privacy concerns, drive ad targeting and thus performance. MMM should help advertisers identify which marketing channels and creatives produce the best returns. Coupled with new ad targeting techniques, MMM will almost certainly improve performance.
Meta’s Robyn, for example, helps advertisers analyze the performance of campaigns across Facebook, Instagram, and other channels. It gauges channel effectiveness and optimizes ad spend based on results.
The era of cookie-less targeting encourages the use of MMM. Some of the most forward-looking, high-budget advertisers are considering alternative targeting methods and new promotional channels. Monitoring those experiments requires complicated multi-touch attribution or MMM.
For example, Google’s Meridian MMM moved beyond standard regression models to a theory called “Bayesian causal inference,” which captures the impact of imprecise marketing actions, such as a social media post.
Personal privacy is yet another reason why MMM appeals to Google, Meta, and many advertisers. The model aggregates data and generally avoids personally identifiable information.
Artificial Intelligence
AI makes MMM relatively faster, more adaptive, and easier to scale.
Improved speeds come first from training the model quickly. The foundational, now-available models are a massive headstart compared to starting from scratch.
Second, AI helps process and clean large, complex datasets from multiple sources such as digital ads, TV, print, and online and in-store sales. The associated algorithms detect seasonality, outliers, and data anomalies, reducing manual work but still requiring data scientists to fine-tune the models.
Regardless, the AI behind Meta’s Robyn dynamically adjusts model variables, improving accuracy automatically. It is thus more adaptable and scalable.
Cloud Computing
Twenty years ago, the rise of web-based marketing produced massive datasets and hefty processing loads. Analyzing that info typically required custom-built infrastructure and expensive data warehouses
These limitations no longer exist thanks to cloud computing advances and affordability. Instead of spending $500,000 or more on MMM software and servers, a company can run Google Meridian in the cloud for a fraction of the amount, perhaps as little as $10,000 a year.
Accurate modeling, however, requires some scale — businesses investing at least $500,000 per year in advertising likely benefit the most. But that could change if MMM becomes available as a service.
