Incrementality testing: the cheapest truth in marketing

Every reporting suite in the world will tell you a channel is working. Only an incrementality test will tell you whether the channel is causing anything to happen.

The question incrementality answers

If we turned this channel off tomorrow, how much revenue would we actually lose? Not how much revenue is attributed to it. How much would disappear.

Attribution is bookkeeping. Incrementality is truth.

The simplest test you can run this month

• Pick a channel you suspect is taking credit for demand it didn't create, often brand search or retargeting.
• Turn it off in two matched geographies for two to four weeks.
• Leave it on in two matched control geographies.
• Compare gross profit, not just revenue, between the two groups.

This is geo-holdout testing. It is unglamorous, it works, and it costs less than a single campaign flight.

What you usually find

• Brand search is 60–90% incremental on net-new customers, much lower on returning ones.
• Retargeting is often 20–40% incremental. The rest would have converted anyway.
• Upper-funnel video is harder to measure but more incremental than its attribution suggests.

Why most brands do not do this

Because the answer is sometimes uncomfortable, and because nobody assigned the test to anyone. Both are fixable. Start with one channel, one quarter, one honest read.