Google Goes All-In on Meridian, Meta Quietly Steps Back From Robyn:
What Platform-Owned MMM Won't Show You

The two biggest free MMM tools just took opposite paths

Marketing mix modeling's 2026 comeback has been driven, in large part, by two platforms giving their own MMM tools away for free: Google's Meridian and Meta's Robyn. Both were released as open-source projects, both were pitched as a way to democratize a technique that used to require a data science team and a six-figure agency retainer, and both, for a while, looked like they were on the same trajectory. We compared the two tools' technical merits and adoption incentives in more detail here.

They aren't on the same trajectory anymore. Over the course of 2026, Google has moved to make Meridian a default, embedded part of its ad and analytics stack. Meta, according to multiple agency and measurement-vendor sources cited by AdExchanger on July 14, 2026, has quietly stopped actively promoting Robyn and scaled back the internal engineering team behind it. One platform is doubling down on owning MMM. The other appears to be walking away from it. Neither outcome should make marketers comfortable relying on either tool as their system of record.

Google's build-out: Scenario Planner, then GA4 360

Google's push has come in two stages. On February 19, 2026, Google launched Scenario Planner, a no-code interface built on top of Meridian that lets marketers without a data science background test different budget allocations and see projected ROI without writing Python. The timing wasn't random - an October 2025 Harvard Business Review Analytic Services report found that nearly 40% of organizations struggle to translate MMM outputs into actionable business decisions, and Scenario Planner was built specifically to close that usability gap.

Three months later, at Google Marketing Live on May 20, 2026, Google went further: it announced that Meridian would be built directly into Google Analytics 360. The integration combines first-party GA4 signals with cross-channel data from Google, TikTok, Pinterest and Snap inside a single model, adds a natural-language interface for scenario planning (a marketer can ask something like "what happens to ROAS if I shift 15% of my YouTube budget to Demand Gen?"), and introduces two new metrics: Attributed Branded Searches, which tracks short-term intent an ad created, and Qualified Future Conversions, a Gemini-powered metric that projects conversions up to six months out.

Taken together, that's a real product. It's also, unmistakably, Google's product - running on Google's data, inside Google's analytics platform, digesting spend data primarily from Google and a handful of other digital ad platforms.

Meta's retreat

Robyn's trajectory has gone the other way. AdExchanger's July 14, 2026 reporting quotes agency and measurement-vendor sources saying Meta has stopped actively pushing Robyn as a strategic priority and has scaled back the engineering team maintaining it. Meta hasn't formally announced a discontinuation, but the signal is clear enough: an open-source MMM tool is only as durable as the platform's interest in maintaining it, and that interest can quietly evaporate.

The structural problem: the model, the media seller and the media are the same company

Set aside the diverging trajectories for a second, and the more important issue is the same for both tools: a platform that sells advertising inventory building the model that grades whether that advertising worked is a conflict of interest baked into the product, not an edge case.

That conflict shows up in two concrete ways. First, the incentive: a model built and maintained by the company selling the media it's measuring has every reason to shade its outputs toward showing that media performing well - even unintentionally, through choices about default priors, calibration data or how the model handles the platform's own attribution signals. Second, and more mundane but arguably more limiting: coverage. Meridian-in-GA4-360 ingests Google, TikTok, Pinterest and Snap. That's a meaningful digital footprint, but it's still exclusively digital. Television, out-of-home, radio and print - the channels that still carry a large share of most brands' awareness-building budget - sit outside the model entirely, or have to be bolted on manually if they're included at all.

For a brand running a genuinely omnichannel plan, that means the flagship free MMM tools from the two largest ad platforms in the world can only ever tell part of the story, by design, because modeling the other part isn't in either platform's commercial interest.

What the industry's own standards body is telling marketers to do instead

This isn't a fringe critique. The IAB's own "Modernizing MMM: Best Practices for Marketers" guide, published in December 2025 in partnership with MMM providers across the industry, explicitly frames vendor-neutral, cross-channel integration - paid, owned and earned media together, calibrated against experimentation and attribution - as a baseline requirement for a modern MMM practice, not an optional upgrade. A model that can only see its own operator's channels, however well-engineered, doesn't meet that bar.

None of this means Meridian or Robyn are bad tools, or that free, accessible MMM is a bad thing for the industry - quite the opposite; both did real work in getting marketers comfortable with mix modeling again after years of over-indexing on multi-touch attribution. But "accessible" and "sufficient as a system of record" are different claims. A model a brand uses to defend its full marketing budget to the CFO needs to treat every channel - the platform's own inventory and everyone else's, digital and offline - on equal footing, with no commercial stake in how any single channel comes out looking.

Independence and offline coverage aren't nice-to-haves - they're the whole point

Google doubling down on Meridian is a net positive for MMM's credibility as a category; it puts real engineering resources behind a technique the industry needs. But it also sharpens exactly the question marketers should already be asking about any MMM vendor: who built this model, what's their commercial relationship to the channels it's measuring, and can it actually see the full media mix - TV, out-of-home, radio and print alongside search, social and retail media - or only the slice that happens to run through its own ad server?

Animo was built around the answer the platforms structurally can't give: a single, always-on causal model that sits outside any one media seller, ingesting offline channels - TV, OOH, radio, print - and online channels together, with no channel getting the benefit of the doubt because it happens to belong to the company that built the model. As MMM keeps consolidating around a handful of dominant tools, that independence is going to matter more, not less.

Sources

• AdExchanger, "Google's Meridian And Meta's Robyn: A Gift To Measurement Or Trojan Horses?" by James Hercher, July 14, 2026: adexchanger.com
• Google, "Enhanced budgeting tools in Google Analytics powered by Meridian" (GA4 360 / GML 2026, May 20, 2026): business.google.com
• Google Blog, "Google Analytics 360 helps you turn data into decisions": blog.google
• Ad Measurement Weekly, "Google Rebuilds Its Measurement Stack at GML 2026: Meridian Moves Into Analytics 360 and a New Predictive Metric Arrives": admeasurementweekly.com
• MarTech, "Google launches no-code Scenario Planner built on Meridian MMM," Feb 19, 2026: martech.org
• MediaPost, "Google Adds No-Code Tool To Its Marketing-Mix Model Platform," Feb 19, 2026: mediapost.com
• IAB, "Modernizing MMM: Best Practices for Marketers," Dec 2025: iab.com (PDF)