What is retargeting?

Why it looks so good

Retargeting posts the best measured performance of almost any digital channel. Conversion rates are high, attributed ROAS is excellent, and cost per acquisition is low.

All of that is true and most of it is a selection effect. Retargeting reaches people who already visited, already showed intent, and were already closer to purchasing than anyone else you could reach. The advertising is present at the end of converting journeys because it is aimed at people whose journeys were converting.

The incrementality trap

The question that matters is not whether retargeted users convert. It is whether they convert *more* than they would have without the advertising.

Holdout tests routinely find that a substantial share of attributed retargeting conversions would have happened anyway. The channel is rarely worthless and is almost always worth considerably less than its attributed numbers claim.

This is the clearest and most expensive example of why incrementality testing exists. It is also the test most often avoided, because the result is unwelcome and the channel has internal advocates.

Doing it well

• Hold out. A random slice of the retargetable pool, compared on total conversions.
• Exclude recent converters. Advertising a product someone already bought is pure waste and mildly insulting.
• Cap frequency hard. Retargeting is the format most associated with feeling followed, and the reputational cost is real.
• Segment by intent depth. A cart abandoner and a homepage bouncer need different treatment and are worth different amounts.
• Expect smaller pools. Identifier loss shrank the addressable audience substantially.