Key takeaways
- Marketing mix modeling (MMM) adoption is accelerating, but only 28% of organizations say they're effective at converting model outputs into action.
- The hardest part of an MMM engagement is the cross-functional alignment, change management, and internal politics that determine whether insights lead anywhere.
- MMM ownership requires marketing, data science, and analytics teams working in lock step.
- Automation of model refreshes makes quarterly cadence sustainable and removes a significant operational barrier to acting on insights.
A recent Harvard Business Review Analytic Services report put a name to something practitioners have been navigating for years: the MMM actionability gap.
Across 547 marketing leaders, the research found that 87% of organizations consider marketing mix modeling (MMM) important for gaining data-driven insights, but only 28% say they're very effective at converting those insights into timely action. Separately, 66.3% of US brand and agency marketers told EMARKETER that aligning marketing metrics to business outcomes is their single top measurement priority.
The infrastructure to do that is getting better. Acting on what it surfaces is a different challenge.
We've spent the last year-plus building and deploying Google Meridian for clients across industries. What the research confirms — and what we've seen play out repeatedly — is that the actionability gap is driven as much by internal silos and budget politics as by technology. Getting that part right is what separates organizations that use MMM to make decisions from those that use it to produce reports.
What's actually preventing teams from acting on MMM results
The HBR report identifies three overlapping barriers to MMM success:
- Technology hurdles: Data quality issues, siloed inputs, and integration challenges with other measurement tools
- Operational friction: Slow internal processes, unclear accountability, and difficulty connecting model outputs to real planning decisions
- Organizational obstacles: Talent scarcity, siloed teams, and low trust in model outputs
Channel teams that have historically owned measurement don't always welcome a model that redistributes credit across the mix. Stakeholders who've built their media plans around multi-touch attribution aren't always eager for a methodology that tells a different story.
The research touches on this when it notes that 34% of respondents cite unclear accountability to follow through on MMM recommendations as an operational obstacle. In our experience, that accountability gap tends to be a symptom of teams protecting what they've built and what they’ve always known.
MMM ownership belongs to more than one team
Successful marketing mix modeling programs require three distinct skill sets, each contributing something the others can't: data engineering, data science, and marketing strategy.
Data engineering
Building and maintaining the centralized data infrastructure the model runs on is a distinct technical skill set that has nothing to do with interpreting marketing outputs. It's also frequently the capability organizations underestimate until they're already mid-engagement, when the reality of getting clean, unified data into a warehouse that Meridian can actually use becomes the primary obstacle.
Data science
Model configuration, iteration, and validation is where the statistical work happens, and where most organizations' sense of "MMM expertise" begins and ends. But building a model and knowing what to do with what it tells you are different skills and conflating them is one of the more common ways organizations set themselves up to close the gap slowly.
Marketing strategy
Marketing practitioners are the people who need to connect model outputs to actual planning decisions, and they're often the last ones brought into a marketing mix modeling engagement. Without this layer, even a well-built model produces findings that sit in a report rather than informing a budget conversation, which is precisely the pattern the HBR research describes.
The HBR report's finding that 45% of organizations cite limited in-house expertise as a barrier is accurate, but the more useful way to think about this is that most organizations are missing at least one of these three capabilities.
|
IN PRACTICE In our Alaska Airlines work, getting the model to a place where outputs could actually drive decisions required cross-functional cooperation across teams that don't typically work together. That meant coordinating on:
-
- Media spend data and normalization across channels
- Brand, organic, and CRM signals
- Control variables like weather, competitive activity, and regional events
That process is what built the internal buy-in that let the model's outputs land. By the time Meridian was running, the right people had been part of building it.
|
Why model refresh cadence matters in MMM and what makes it feasible
The HBR report notes that leaders update or refine their model design quarterly or more often, and that frequency correlates with better outcomes. What it doesn't address is what makes that cadence actually sustainable.
A quarterly refresh means pulling updated data, re-running the model, validating outputs, and getting stakeholder alignment on what changed — every 90 days, on top of everything else those teams are managing. Without automation, that operational burden is real enough that refreshes slip, and a model that isn't being updated regularly starts producing insights that lag behind the market conditions teams are actually navigating.
This is the part of the marketing mix modeling conversation we think is undersold. Reducing the ongoing operational burden is what keeps the model useful past the initial build. Two recent additions to the Meridian ecosystem address this directly:
- Meridian Studio is an enterprise platform built on Google Cloud that helps organizations scale and tailor large volumes of marketing mix models. This makes the refresh cycle a manageable operation rather than a full team engagement.
- Meridian GeoX is an open-source geo lift tool that runs transparent, publisher-agnostic incrementality experiments to calibrate your model and validate its outputs against real-world results. This should help close the loop between what the model predicts and what's actually happening in market.
Additionally, Google Query Volume feeds live consumer search behavior directly into Meridian. That means Meridian is reading actual in-market demand signals in near real time, rather than just historical data.
The direction this is heading goes further than automation of refreshes. Agentic solutions that can actively monitor model outputs, flag when conditions shift, and surface recommendations without waiting for a human to pull the data are already in development and closing the actionability gap even further.
What determines whether MMM insights actually change decisions
When outputs surface findings that contradict what teams have assumed — about which channels are driving results, about where budget is actually earning its keep — the response isn't always "now we know and we'll act on it." Sometimes it's genuine uncertainty about what to do next. Sometimes it turns political. Often it's both.
What we've found is that the organizations that close the actionability gap most effectively have established clear ownership before the model was built. Without that, insights tend to fall by the wayside — not because the model failed, but because no one had the internal backing to act on what it found.
If you're navigating an MMM implementation or trying to figure out why model outputs aren't driving the decisions they should, we'd love to talk.
