Your platform numbers say a channel is delivering strong ROAS. Your MMM suggests you’ve probably overspent there. An incrementality test indicates a positive lift. Which number should determine where the next dollar goes?
The instinct is often to deem one of these as the ‘Source of Truth’, but in every likelihood all of these models are wrong to a certain degree. Each measurement method is giving you an answer, but most likely to different questions. The more useful approach is to start with the decision you’re trying to make, then establish which evidence should have the authority to make it.
Begin by giving each measurement method a job
At Brainlabs, we think about measurement as a connected system with different methodologies operating at different levels. MMM informs strategic allocation, helping determine where investment should broadly move across the media mix. Incrementality validates causality, testing whether an intervention actually created an outcome that wouldn’t otherwise have happened. Attribution and platform signals support day-to-day, mostly in-channel, optimization, helping teams make faster decisions within those broader investment guardrails.
The same channel can look very different through each lens without one of them necessarily being wrong. A platform can be extremely good at identifying people likely to convert and therefore report strong returns, while an incrementality test reveals that many of those customers would have purchased anyway. Both findings are useful. They simply shouldn’t be allowed to make the same decision.
Once you establish those decision rights, the question changes from “Which number is right?” to “Which number is right for the decision I’m making?”
Don’t be afraid of measurement failure
If MMM suggests a channel has significant headroom but an incrementality test finds limited causal impact, averaging the two results until they tell a more comfortable story won’t help. The discrepancy is telling you there’s something worth investigating. Perhaps the experimental window isn’t capturing longer-term effects. Perhaps the test design or audience introduces a limitation. Or perhaps an assumption in the MMM needs to be revisited.
The same applies when platform ROAS looks exceptional but incremental performance is weak. That gap can reveal something important about the role the channel is actually playing: it may be highly effective at finding and converting existing demand without creating as much additional demand as the attributed return suggests.
This is where multiple measurement methods become more valuable than one. Each can expose assumptions the others can’t see. Use measurement failure to identify where your understanding of performance needs to get sharper.
Make your measurement methods learn from each other
The next step is making sure that insight doesn’t stay trapped inside whichever analysis produced it.
Suppose MMM identifies potential headroom in a channel. Rather than immediately making a large allocation change, that finding can become a hypothesis for experimentation. A controlled test establishes whether increasing spend genuinely creates incremental growth. That causal evidence can then feed back into future planning, strengthening or challenging the assumptions that informed the original recommendation.
The same loop should continue into execution. Strategic measurement establishes where investment should broadly go, experimentation pressure-tests the decisions that matter most, and attribution and platform signals help teams optimize within the resulting guardrails.
So rather than three competing versions of performance, you get a feedback loop:
MMM identifies the opportunity, experimentation tests the assumption, the learning improves future planning, attribution guides execution within those boundaries.
The value comes from the connection between them. Each methodology makes the next decision better informed.
Decide disagreements before they become political
This becomes particularly important at enterprise scale, where different teams naturally gravitate toward different evidence. Performance teams have platform data, analytics teams may own MMM and experimentation, while finance has its own view of revenue and return. Without an agreed hierarchy, measurement disagreements can quickly become organizational ones.
The answer isn’t choosing one metric that everyone has to use. It’s agreeing upfront on which evidence governs which type of decision and what happens when signals conflict.
That means defining the rules before the numbers arrive. Which methodology determines strategic budget allocation? When should an incrementality test challenge an existing assumption? Which signals can teams use to optimize independently, and when does a decision become significant enough to require a different level of evidence?
This is what turns measurement from a collection of reports into a decision-making system. Teams can still look at different metrics because they have different jobs to do, but everyone understands how those metrics ultimately connect back to investment decisions.
Stop looking for one measurement truth
Modern marketing is too complex for one methodology to answer every question well. Trying to force MMM, incrementality and attribution into agreement can remove the very differences that make having multiple methods useful in the first place.
Start with the decision. Give each methodology a clear job. When the evidence disagrees, investigate what the gap is telling you. Then feed what you learn back into the system so the next decision starts with better information than the last.
The goal isn’t to find the one number everyone can agree on.
It’s to know which number deserves to make the decision.
