Marketing Teams Are Rebuilding Trust in MMM Outputs Through Cross-Functional Context

The defensive era of MMM is closing. When COVID forced 40 to 50 percent budget cuts across marketing organizations, teams reached for marketing mix modeling because they needed a defensible way to decide what to cut. The math gave them cover. The output gave them slides. What it rarely gave them was a real shift in how budgets were set, because the models were often deployed downstream of the decisions they were meant to inform. That pattern is now breaking, with marketing leaders increasingly recognizing that MMM only delivers strategic value when it sits inside a broader measurement framework that pairs modeling with experimentation, context, and cross-functional trust.

Architecting modern solutions to this problem is Manoj E., Partner and CMO of Garner Business. He spent years building MMM solutions for in-house brand teams at large vendors across SaaS, CPG, and energy before moving into consulting, where he now helps organizations design the foundational frameworks that drive growth. His approach treats MMM as one instrument inside a wider system, which allows him to be direct about where it falls short on its own.

"Channels like SMS often get under-credited simply because MMM doesn't have clean, structured visibility into how to value them," he says. The gap he identifies runs deeper than a single channel. It points to what MMM can and cannot see, and the human work required to close the difference.

How MMM became a tick-box exercise

In its second wave of adoption, MMM moved fast and shallow. Teams under pressure to defend spend looked at vendor reputation, slot in the Gartner or Forrester quadrant, and presentation polish. The underlying modeling approach received far less scrutiny. Once the solution was in place, leaders often used it to confirm what they had already concluded. "All through the initial days, tools like MMM were being used as a tick-box item. People would largely just blindly use MMM findings to justify what they've already decided," Manoj explains.

Expensive solutions ended up producing expensive validation. The strategic potential of the tool, the ability to compare incremental impact across channels and reshape future allocation, sat largely unused. What is changing now, in his read, is that marketing teams have begun asking different questions before they buy. "A lot of teams that I work with right now are understanding that there needs to be a more foundational framework, and then the need for tools comes from there, whether it is MMM or attribution or experimentation."

The order of operations matters. Framework first, then tool selection. Without it, every vendor decision is a bet that the organization will know what to do with the output once it arrives.

The trust problem is bigger than the math

Even when MMM is technically sound, it can be ignored. Manoj points to the trust gap that opens between data teams, marketing leaders, media agencies, and brand teams the moment a model produces a result that contradicts intuition. CPG performance marketers tend to be the most disgruntled, because MMM frequently shows that traditionally performance-tagged campaigns deliver less incremental lift than the attribution dashboards suggest. "If they don't have the trust, then the solution falls flat on its face," he says.

His response is to build context into the modeling process from the start. Sales leaders, channel partners, and retail teams carry years of on-the-ground knowledge that no model can replicate. Capturing that context, even when it lives in emails and unstructured comments, becomes part of the work. "How do we quantify the context that these teams are giving? It could be offhand comments, it could be just something written over email, not necessarily structured data. The idea is how do we structure all of this context, which is beyond numbers."

Top-down sponsorship is the other half of the trust equation. Without it, data collection slows and cross-functional cooperation stalls. He sets expectations at the KPI stage, before the model produces any output that someone will need to defend. "Especially in CPG, there are lots of surprises that come up. Things like on-ground work, sponsorship, grassroots-level marketing. So we try to set those expectations at the beginning."

Why SMS keeps getting under-credited

The framework problem becomes concrete when leaders look at owned, high-intent channels. SMS is one of the highest-margin channels in the modern marketing stack, with engagement and conversion metrics that rival or exceed paid digital. It also tends to get under-credited in MMM outputs. "Any place where MMM doesn't have a clear look at what kind of metrics could come into play, there will be some kind of under-crediting happening," Manoj notes. 

The asymmetry is rooted in data hygiene. CPG organizations tend to retain data across structured and unstructured systems, which is one reason MMM works well in that vertical. Other industries, and other channels, do not. Model choice compounds the bias. "Bayesian models really need more context to manually set benchmarks. If marketing teams think SMS has a higher value, then the model will show SMS as the higher level. That's my personal grouse with the Bayesian models. They put a lot of focus on marketing context, whereas marketers are not programmed yet to put that context in numbers."

The implication for leadership is direct. When an MMM output suggests an owned channel is underperforming relative to what real-time engagement data shows, that conflict is a signal to investigate the inputs rather than immediately acting on the outputs. 

AI is not the rescue

MMM and attribution have used machine learning and text analytics for years. The recent generative wave has mostly added an insights layer that, in Manoj's view, does not yet do a great job. What has changed is how mature teams use general-purpose AI in their day-to-day workflows. "Most of their data teams were using Claude skills as part of their daily drivers. They were pretty comfortable using a tool like Claude to actually analyze any data output that the model is throwing out, and not really depending on that tool or a measurement tool itself to have it baked in." The human interpretation layer is getting faster and more rigorous without requiring the underlying measurement tool to change. That capability matters most when the framework is right.

Some of the mature B2B clients Manoj works with skip MMM entirely, particularly traditional shipping and oil and gas, where leaders already trust the impact of their brand work and need a different kind of closed-loop measurement. The decision about which tool to apply where flows from the framework, rather than the other way around.

For marketing leaders setting next year's budget, the call is to stop asking whether MMM is right or wrong about a given channel and start asking whether the surrounding system can be trusted to tell them when it is. Experimentation becomes the feedback loop that makes the next model smarter, and the discipline that turns MMM from a defense of last year's spend into a credible input on next year's bets. "MMM is not a one-stop solution," Manoj asserts. "Neither was attribution modeling a few years back. It's about building that framework. A great way to go about it is having experimentations running along with it in parallel. That gives you an idea of whether you're under-crediting or over-crediting, and that then gets put in as context in the next model."