For a decade, the marketing mix model has been an annual event: a data science team disappears for six weeks, comes back with a deck, and the budget fight starts over. Google’s latest overhaul of Meridian, its open source MMM, points at a different future, where the model itself flags its own bad data, resolves the errors it can, and updates in something closer to real time. The shift is not that Google built a better spreadsheet. It is that measurement is turning into infrastructure a marketing team has to maintain, not a report it waits to receive.
What Google Actually Changed
The update, announced on Google’s own blog by Nipoon Malhotra, VP of Ads Analytics, Insights, and Measurement, bundles three changes into Meridian. First, new agentic capabilities audit data quality, resolve errors, and guide model building in real time, work that previously required a specialist to catch manually before a model’s output could be trusted. Second, the back end got faster, which matters more than it sounds: a model that takes days to rerun does not get rerun when a data pipeline breaks, it gets used anyway. Third, Meridian can now ingest brand signals such as Branded Google Query Volume directly, so upper-funnel formats like TV and out of home get a path into the same causal framework as search and social.
Alongside the Meridian changes, Google also took Meridian GeoX, its library for causal geo-experiments, to global general availability after a beta. GeoX lets a team run an independent incrementality experiment or feed the results back into the mix model to sharpen its accuracy, which is the closest thing performance marketing has to a control group. And the Data Manager API, the pipe that feeds first-party data into all of this, is now built on the IAB Tech Lab’s Event and Conversions API standard and works across major ad platforms rather than one at a time.
Why an Audit Layer Changes the Argument
An open source MMM used by hundreds of advertisers functions as a de facto industry reference point the way a widely adopted file format does. When Google adds an agentic error-checking layer to that reference point, it resets what “credible measurement” looks like everywhere else, including in tools that have nothing to do with Google. A brand’s own data science team can point to a broken pipeline and say, correctly, that catching it required expertise the company did not have. That excuse gets weaker when the industry’s most-copied model catches the same class of error automatically.
The brand-signal change matters for a different reason. Upper-funnel spending on TV and OOH has always been the easiest line item to cut, precisely because it has been the hardest to hold accountable to a number. Folding branded search volume into the same causal model used for performance channels does not make brand spend as measurable as a paid search click, but it puts both on a model that at least uses the same auditing standard. Google reports that advertisers connecting offline and app data through Data Manager see an average 26% increase in incremental ROAS, and that enhanced conversions in Search deliver an average 11% conversion increase, numbers pulled from the same measurement stack the Meridian updates plug into.
The Data Manager API change is easy to skip past, but it is doing quiet structural work. Building the pipe on the IAB Tech Lab’s Event and Conversions API standard means a first-party data connection built once can be reused across ad platforms instead of rebuilt for each one, which is the kind of unglamorous plumbing that determines whether a mid-size marketing team can actually keep its data foundation current or falls behind every time a platform changes its integration requirements. Google also reports that Google Tag Gateway users see an average 14% conversion uplift and that Demand Gen campaigns see uplift above 20%, figures that only mean something if the underlying data connection feeding them is trustworthy, which is exactly what the new agentic auditing is meant to police.
What It Means for the Marketing Leader
The practical order of operations has flipped. It used to be reasonable to commission a big mix model and hope the underlying data was clean enough to trust its output. Now the audit comes built into the model, which means a leaker or broken tag shows up as a flagged error instead of a silently wrong coefficient three months later. That is a reason to push vendors, not just Google, on whether their measurement stack does this kind of continuous checking or just produces a score and hopes.
It also changes how an upper-funnel budget gets defended. Measurement has been racing to catch up with faster-moving channels for a while now, and a causal geo-experiment run through GeoX is a stronger opening position in a budget conversation than a brand-lift survey, because it is designed to isolate incremental effect rather than describe correlation. Marketing leaders heading into 2027 planning should treat an audit trail, not just a lift number, as the deliverable they ask measurement vendors for.
The Agentic Layer Is the Real Story
None of this is unique to marketing measurement. Ad tech has spent the past year building standard ways for agents to plug into existing systems, and Meridian’s update is the same pattern applied to attribution: an agent does not replace the analyst, it does the first pass of the analyst’s least favorite job, which is finding out why the numbers look wrong before anyone builds a strategy on top of them. The publications that treat this as a Google feature update will miss the larger point. The mix model is becoming a system that has to be maintained continuously rather than commissioned periodically, and the teams that adapt their planning calendars to that reality first will spend the next budget cycle arguing from an audited number while everyone else argues from a deck.
None of this requires a company to use Meridian specifically. The standard it sets, continuous data auditing built into the model rather than bolted on afterward, is now the bar every other mix-model vendor gets measured against, whether they are ready for the comparison or not. Marketing leaders evaluating a measurement partner in the next planning cycle have a new, concrete question to ask: show the audit trail, not just the score.
Source: Google
