The incrementality test worked. One channel, clean results, defensible numbers for the CFO.

Then someone asks: "Can we do this across our entire retail media spend?" and that's where it falls apart.

The problem isn't the test. It's what the test can't see.

Over half of US marketers are running incrementality tests. Another 36% plan to invest more next year. The methodology isn't the mystery anymore; teams understand holdout groups, control groups, and statistical significance.

The problem is that most incrementality testing measures performance within platforms when the actual question is about performance across your entire growth system.

Your Meta test worked because Meta controls the environment. Clean audience split, isolated exposure, measurable lift, but your brand doesn't run on Meta alone.

It runs on a fragmented system where Meta prospecting might drive Amazon purchases three weeks later. Where retail media might just be stealing from organic search. Where your channels are cannibalizing each other, and nobody can prove it.

The questions one clean test can't answer

Here's why 43% of marketers cite "applying incrementality across ad types, targeting methods, and retailers" as their top barrier.

It's not that they don't understand the methodology. It's that the test that worked perfectly on Meta can't answer the questions that actually matter:

Is paid media creating new demand or just capturing people who were already going to buy from you?

Which channels are actually building your pipeline versus harvesting existing demand?

Your Meta incrementality test shows 3.8x iROAS. Your Amazon test shows similar lift. Both platforms report strong incremental performance. And yet when you look at total business growth, the math doesn't add up.

That's because platforms measure incrementality within their walled gardens. They can tell you if their ads drove more conversions than no ads. They can't tell you if those conversions would have happened anyway through a different channel.

What actually works at scale

The teams getting this right aren't running better individual tests. They're asking different questions at different altitudes.

At the portfolio level: Which channels are building demand versus capturing it? Marketing Mix Modeling shows patterns your platform dashboards can't see, like Meta prospecting creating Amazon conversions three weeks later.

At the market level, did the campaign actually move the needle, or did we just shift the timing? Geo tests answer this by isolating geography rather than audiences.

At the execution level: Within this channel, what's actually incremental? Platform testing works here, but only after you've validated the channel strategy.

When these contradict each other, that's not measurement failure. That's useful tension showing you where your assumptions are wrong.

The gap between testing and infrastructure

Incrementality testing isn't hard because the methodology is complex. It's hard because businesses don't operate the way measurement frameworks assume they do.

Platforms want you to measure within their walls. Your customers move between them. Attribution wants to assign clean credit. Your channels overlap and interfere with each other. Single tests want controlled environments. Your media mix is chaotic, fragmented, and changes weekly.

The teams that figure this out stop treating incrementality as a measurement project and start treating it as strategic infrastructure. They build systems that answer cross-platform questions. They align Finance and Marketing on what "incremental" actually means. They accept that perfect measurement is impossible, but directionally correct measurement beats dashboard theater.

Most teams aren't there yet. They're stuck between "we ran one clean test" and "we have no idea if our media spend is actually incremental."

You can have perfect incrementality measurement on every individual platform and still be completely wrong about what's driving business growth.

The question isn't whether incrementality testing matters. It's whether you're measuring the right thing.