Google is making several updates to Meridian, its open-source marketing mix modeling tool, with a bigger focus on incrementality and causal measurement.
The biggest update is the global launch of Meridian GeoX, which moves from beta to general availability.
Google is also adding new capabilities to Meridian that can help marketers build models and account for longer-term brand effects.
The updates give advertisers more ways to measure how their media investments contribute to business results beyond what platform attribution can show.
Meridian GeoX Is Now Available Globally
Google first previewed Meridian GeoX in May as an open-source tool for running geographic incrementality experiments.
Now, GeoX is generally available globally within Meridian.
GeoX allows marketers to run experiments across geographic areas to estimate the incremental impact of advertising. The goal is to determine how much of an outcome happened because of the media investment.
It can also be used across advertising platforms, so marketers aren’t limited to testing Google media.
A useful example could be if an advertiser wanted to use geographic testing to evaluate campaigns running on another platform. It could also test media running across several channels.
That makes the global launch more useful for advertisers trying to understand performance across their entire media mix.
GeoX Results Can Help Calibrate Meridian Models
GeoX also ties directly into Meridian’s role as a marketing mix modeling tool.
Marketers can incorporate incrementality results from GeoX into their MMM, giving the model another source of evidence to work from.
MMM relies on historical data to estimate how different investments contributed to business outcomes. Those estimates can get murkier when several channels move at the same time or outside factors influence performance.
Geo experiments can add more confidence around specific investments by testing whether they produced incremental results. Google says those findings can then be incorporated into MMM to help improve accuracy.
That creates a closer connection between modeling and experimentation. Meridian can provide a broader view of performance, while GeoX adds causal evidence from individual tests.
GeoX also gives marketers another option when user-level experiments aren’t practical. Geographic testing can create treatment and control areas without relying on individual user tracking.
Bringing more experimental data into Meridian is only one part of Google’s latest changes. Google is also adding new tools aimed at making the process of building and working with those models easier.
Agentic Tools Will Help With Meridian Model Building
Google is also adding agentic capabilities to Meridian to assist marketers during the model-building process.
The tools can audit data quality, help resolve errors, and provide guidance as a model is being built. Google says the assistance will happen in real time.
That could take some of the manual troubleshooting out of the process, especially when issues with the underlying data need to be identified before an analysis can move forward.
Google is making changes to Meridian’s backend as well, with the goal of running analyses faster and more efficiently.
The updates don’t change the inputs marketers need to make a model useful. Teams still need to determine which business and media data belongs in the model and how much confidence to put in the results.
Google’s approach appears to focus more on helping marketers work through the technical process. With GeoX adding another source of experimental data, those tools could become more useful as marketers bring additional signals into Meridian.
Meridian Can Now Account For More Brand Signals
Meridian models can now incorporate relevant brand signals, including Branded Google Query Volume.
Google says these signals can help marketers account for the longer-term effects of brand investment. A TV commercial or OOH campaign, for example, may increase interest well before that demand turns into sales.
Branded query volume gives marketers another signal to examine during that period. It can help show whether branded demand changed alongside upper-funnel media investment and how that activity fits into the larger model.
However, there are some caveats to how marketers interpret that data.
An increase in branded searches alone doesn’t establish that a campaign caused future sales. Competitor activity, seasonality, promotions, news coverage, and other factors can influence branded demand at the same time.
Meridian can account for brand signals alongside the other business and media data included in the model. That gives marketers more information to work with when evaluating investments whose impact may develop over a longer period.
The addition also broadens the types of questions marketers can bring to Meridian as Google builds out its measurement tools beyond immediate campaign performance.
What This Means For Advertisers
GeoX could make Meridian more approachable for marketers trying to get more from MMM, especially when those findings influence larger budget decisions.
One of the challenges with MMM is explaining how much confidence to put behind the results. Models rely on historical data and statistical assumptions, which can make it harder to defend a major budget shift based on modeling alone.
GeoX also gives marketers a way to add experimental evidence to that conversation.
If a geo experiment supports what the model is showing for a specific channel, marketers have another piece of evidence to bring into budget discussions. That could also make MMM findings easier to communicate to executives who aren’t involved in the methodology behind the model.
There is still a fairly high barrier to using these tools well.
Meridian in itself is free and open source, so there isn’t a software licensing cost to start using it. But, there are still costs associated with the people, data, infrastructure, and media required to build models and run experiments.
Google recommends GPU resources for Meridian because the modeling can be compute intensive. GeoX also has specific data requirements, including daily time-series data and enough geographic variation to design a reliable experiment.
The media cost of experimentation needs to be considered as well. Depending on the test, marketers may need to increase, reduce, or hold back spend across selected markets to create the conditions needed to measure incrementality.
That puts some limits around who can take advantage of GeoX today.
For advertisers with enough scale and data, though, the combination could change how MMM gets used internally. Instead of presenting modeled results as the final answer, teams can use experiments to test areas where they want more confidence before making larger investment decisions.
Where Meridian And GeoX Go From Here
Meridian has already given marketers another way to look at performance across channels that don’t fit neatly into platform attribution.
GeoX adds another layer by giving marketers a way to test some of what their models are telling them with real-world experiments.
When MMM is informing a significant budget shift, having an experiment to support those findings gives marketers more to work with than the model output alone.
It could also help when explaining those decisions internally. A leadership team may not need to understand every assumption behind an MMM, but showing that an incrementality test produced similar findings can make the recommendation easier to support.
There will still be cases where the model and experiment don’t line up. Those differences may be just as useful, especially if they lead marketers to question an assumption or take a closer look at how a channel is being valued.
As more advertisers start using GeoX, it will be worth watching how often those experiments confirm what their models were already showing and where the results tell a different story.
Featured image: courtesy of Google
