Is Google’s MMM platform Meridian a gift to marketers or another dead end?

I will answer that question and more, but first. Say that you sell candy bars. You change the price that you charge retailers. You run various ads over several channels. You move them to different shelves in supermarkets and corner stores. You know that local economies have varying levels of strength. So, which variable – or which several – affected sales volumes and profits? And how?

Now, imagine that you sell B2B enterprise software. A sales cycle might take years. People might see you at a conference. They might have searched Google for your type of product. They might have read the sales collateral and enablement material on your website. They have talked to salespeople. So, what gets the “credit” for an eventual sale?

The likely answer?

Many or most of those things – to one degree or another.

For a generation now, “digital” and “performance” marketers have fallen for the lie that the “attribution” that we see in online analytics dashboards means something useful. But no one buys anything because of one ad that they see one time. The real world does not function in a linear, neat, and straight line.

In fact, my hours-long effort to get Oasis reunion tickets showed Twitter as the “source” of my purchase despite the fact that I have wanted to see the band for around thirty years.

So, what should marketers do?

As I wrote in a prior column that included a lengthy interview with Les Binet, the former group head of effectiveness at Adam&EveDDB who recently became a private consultant, they should enter the world of marketing mix modeling (MMM) – also called econometrics.

And now, Google has joined that world as well.

The major MMM players today

Basically, MMM does advanced statistical analysis on numerous different variables – anything from ad campaigns to different pricing to the weather to economic factors – to see what factors have the highest correlations with greater sales and profits over the long term. And that will tell you how to set those variables – such as the ones in the B2C and B2B examples above.

It’s not just me. More and more people are seemingly becoming underwhelmed by illogical attribution models and inaccurate analytics dashboards based on website scripts that are easily stopped by ad blockers and GDPR-mandated cookie consent buttons. As a result, an increasing number of analyst and analytics businesses are now offering MMM.

For this column, Gartner sent me the firm’s Magic Quadrant for Marketing Mix Modeling, a November 2024 report that evaluated and rated the major companies in the category. The Leaders were Analytic Partners, Ipsos MMA, OptMine, and TransUnion. The Visionary was Ekimetrics. The Niche Players were Circana, Nielsen, Kantar, Keen Decision Systems, and Fractal Analytics. There were no companies in the Challengers quadrant.

“The popularity of MMM solutions for marketing measurement has risen in recent years as technology, regulatory and consumer perception changes have limited the availability and usage of individual-level advertising data,” Gartner’s report states. “According to the 2024 Gartner Marketing Analytics and Technology Survey, 64% of senior marketing leaders have adopted MMM solutions.”

The firm continues: “High utilization of MMM was associated with marketers’ ability to prove the value of their function and receive credit for contributions to business outcomes. B2C marketing leaders who reported MMM utilization were twice as likely to be able to prove value and receive credit than leaders who said they did not use MMM.”

Google joins the MMM crowd

Google recently announced that its open-source MMM, Meridian, is now available along with a partner program consisting of companies trained and certified on the platform. (Meta released its own open-source MMM, Robyn, a few years ago. It did not respond to a request for comment for this column.)

“As the privacy and regulatory environment continues to change, given their aggregated nature, MMMs are once again becoming an increasingly important part of an effective, comprehensive measurement strategy. They help businesses understand ROI and validate that information across all of their marketing channels in a privacy-centric way,” Harikesh Nair, senior director of data science at Google, told me in an interview for this column.

He continued: “Our goal with Meridian is to help advertisers understand their marketing efforts but also contribute to the industry as a whole by introducing new, innovative methodologies for advertisers to be able to measure against. That’s why Meridian is an open-source model.”

What makes it different? First, Nair said that Meridian replaces “impressions” with “reach” and “frequency.” I wholeheartedly support this because one ad “impression” has always just been one web browser making one request from one server somewhere to load an advertisement. It has never had anything to do with human eyeballs – and it has been the source of a lot of online ad fraud.

Second, people can add incrementality experiments from any channel for which there is data. Third, Nair said Meridian can measure paid search more accurately by breaking down organic-driven demand. Fourth, companies can build on top of Meridian – because it is open source – to create their own solutions.

How can marketers use Meridian?

Forrester senior analyst Brad Haag recently wrote that “digital agencies adept at campaign measurement but lacking an MMM solution will benefit from using Meridian to create solutions for clients. But for brands, the calculus is a little more complicated.”

Brands, he continued, could benefit the most if they have a significant in-house data team, a history of maintaining geo-level marketing, sales, and competitor data, and experience with incrementality testing and ROI calculation.

My note: The point about having historical data is critical. MMM needs at least two to three years of data to reach statistically significant conclusions – and that is one reason for the stubborn refusal of attribution and analytics dashboards to go away. Smaller companies and startups often do not have two to three years of data, so they must rely only on services such as Google Analytics.

Now, this is where I owe Google an apology.

In my columns and speeches, I am frequently critical of tech companies such as Google and Meta for what they are doing to the marketing industry specifically as well as the world in general. As a result, I immediately assumed the worst – that Meridian would be a self-serving tool that would tell advertisers how best to allocate spend on Google’s platforms for the company’s own benefit – as many others have too. I thought it would suggest optimizing spend across areas such as paid search and YouTube but ignore others such as linear television.

But in this instance, I was wrong.

See the generic “channels” shown above. Think of Meridian as a neutral calculator that runs whatever numbers that people import from anywhere.

“Meridian is not just for measuring across Google. The tool helps marketers measure across any channel the customer has data available, such as Google Search, linear TV, radio, and other digital/offline media,” Nair said.

My other concerns were Google’s potential business strategy and data collection practices. Typically, open-source software companies that are for-profit have a basic, publicly available product and then make money by doing things such as selling a business version or offering customer support.

What is Google’s goal with Meridian?

Is Google’s goal to collect the data of everyone using a platform for free and then sell it to advertisers? (As Google and Meta often do.)

I posed these questions to Nair. Here are his responses in full. Take from them what you will.

Nair on how Google financially benefits from Meridian: “Transparency is essential and helps us truly deliver on innovation while bringing the industry along. By open-sourcing the methodology, advertisers can understand the results they are getting through the MMM, choose to build on the MMM or use parts of it that best suit their goals. All methodology is open-sourced through a code library and also published in public research papers, giving the industry transparency into how we’re arriving at the end output.”

Nair on what data Google collects from Meridian users: “Our MMM data platform enables advertisers who want to ingest their campaign performance data across Google properties (ie, Search, YouTube, display, etc.) into any MMM, including Meridian. Advertisers who use Meridian control their own data, and Google doesn’t have visibility into their datasets unless we are helping them to provide Google-specific inputs.”

What existing MMM companies think of Meridian

Still, Meridian does have its detractors. Perhaps not surprisingly, Nancy Smith, president and chief executive of Analytic Partners, was dismissive of Meridian and Meta’s Robyn.

“Open-source code like Meridian and Robyn demonstrate that large platforms have an understanding of the importance of econometric modeling and MMM. These open-source options may be useful for novice analysts, such as students, who want to experiment with one type of econometric model,” she told me.

Smith added: “We have spent the last 25 years building upon the intelligence and capability to support multiple industries and their varied commercial decisioning needs. We are a trusted partner and have delivered measurable and proven value for our customers over the long-term. Meridian and Robyn offer a look under the hood for novice analysts and provide Meta and Google a view into how users or brands may interact with their code.”

For those who want to go deeper into the math and statistics, Aryma Labs, a software company in India, has an in-depth exploration of Meridian on Substack that, in part, criticizes Google for using Bayesian methodology.

And we should not forget that MMM itself as a whole has skeptics. None other than Tom Goodwin has called it “complete nonsense” for reasons including that we cannot quantify creative effectiveness in statistical models.

The future of marketing mix modeling

Regardless of what one may think of Meridian, the entrance of new players into any category will always have an effect.

Matt Wakeman, senior director analyst in the Gartner Marketing Practice, told me that he sees three trends surfacing: increased confusion as new companies enter the MMM market, a software-led shortening of the time needed to do analysis, and more MMM vendors offering to help chief marketing officers sell the practice to the executive suite.

“Two shifts have led to a surge of companies offering MMM solutions: third-party data deprecation and open-source software packages that purport to automate MMM,” he said. “For prospective buyers, new potential offerings can be found from advertising agencies, professional services analytics firms, SaaS multi-touch attribution companies that pivoted to MMM, and SaaS startups focused on MMM and traditional SaaS vendors. This dramatic increase creates uncertainty for buyers trying to understand the differences between providers.”

For me, the main takeaway is that attribution is not the way to measure results, and marketers should never have thought that it was. Online dashboards have always been inaccurate, and Google’s rollout of Google Analytics 4 in 2023 as a replacement for Universal Analytics just made everything worse.

My unverified hunch: Google has always known all of this, but it could never say so explicitly because it would make it look as though it was deceiving marketers with faulty products for decades.

I think we can all agree with what Analytic Partners’ Smith told me and that it is a good outcome: “Meta and Google’s entry into the MMM space with open-source code is validation that MMM is the future for marketing measurement. This is also a signal that the industry is shifting away from attribution methods.”

The Promotion Fix is an exclusive, long-running column for The Drum, contributed by global keynote marketing speaker Samuel Scott.