Notes · The numbers · May 15, 2026

Incrementality, explained for founders.

Here's a question that should keep you up more than your ROAS does: would that sale have happened anyway?

That's the whole subject. Everything else about incrementality is detail.

The problem: platforms grade their own homework.

When Meta or Google reports your ROAS, they're claiming credit for sales their ads "touched." But touched is not caused. Your most loyal customer was going to buy this month regardless. She saw a retargeting ad on the way to the checkout she was already walking toward, and the platform booked the sale as its win. Multiply that across thousands of orders and your dashboard ROAS can look brilliant while a chunk of that spend bought you nothing you weren't getting anyway.

This is why brands cut a "high-performing" retargeting campaign and revenue doesn't move. The ads weren't driving sales. They were standing next to sales, taking a bow.

The idea: measure what changes when you turn it off.

Incrementality flips the question from "what did the ad touch?" to "what happened differently because the ad ran?" The cleanest way to know is the same way medicine knows if a drug works: hold some people out. Show the ads to one group, not to another, compare. The difference between the groups is the incremental effect, the sales that genuinely would not have happened without the spend.

In practice that's run a few ways. Holdout tests split audiences and deprive one group. Geo tests do it by region: turn spend off (or up) in some cities and not others, and watch what actually diverges. Crude, honest, and hard to argue with. Even a simple pause test, switching a channel off for two weeks and watching blended revenue, tells you more than a quarter of dashboard screenshots.

iROAS, in one sentence.

iROAS is just ROAS recalculated on incremental revenue only: not "revenue the platform claims," but "revenue the test proved the spend caused, divided by the spend." A channel can run a platform ROAS of 4 and an iROAS of 1.2. That second number is the one your P&L feels.

When it matters, and when it doesn't.

Honesty cuts both ways here. If you're spending $20k a month, formal incrementality testing is probably overkill; your blended numbers, watched properly, will tell you most of the truth. The discipline matters more as spend scales, because the gap between claimed and caused gets expensive. The principle, though, applies at every size: treat platform attribution as a witness with an interest in the case, not a judge. Cross-examine it against your blended revenue, your contribution margin, and the occasional deliberate test.

What to ask whoever runs your growth.

Three questions sort the operators from the screenshot merchants. Which of our channels have we ever actually tested for incrementality? What would we expect to happen to revenue if we paused this campaign for two weeks? And when the platform number and the blended number disagree, which one do we act on? If the answers are confident and specific, good. If the answer is a dashboard, you have your answer too.

← Back to Notes