Remarketing looks brilliant and is usually overstated

Remarketing consistently reports excellent performance. It targets people who already visited, already showed intent, and were in many cases already coming back.

Then it takes the credit.

The attribution problem

Someone visits, leaves, sees a remarketing ad, returns and converts. The ad is credited. But a proportion of those people would have returned regardless — through direct traffic, a bookmark, or a branded search.

That proportion is invisible in standard reporting, and it is frequently large. The narrower and more recent the audience, the larger it tends to be: a seven-day cart abandoner list is mostly composed of people with high intrinsic likelihood of returning.

How to find out

A holdout. Exclude a random share of the audience — 10–20% — from remarketing, and compare conversion rates between the held-out group and the targeted group.

There is no one-click holdout button; you build it by splitting the audience and excluding one half, via campaign experiments or a duplicated campaign with the audience excluded. Few accounts do it, partly because it is fiddly and partly because the answer is often unwelcome.

The result gives you an incrementality rate: the share of remarketing conversions that genuinely would not have happened. That number, rather than reported cost per acquisition, is what should drive the budget.

Where remarketing genuinely earns its place

Long consideration cycles. Where the decision takes weeks and involves comparison, staying present is real value.

Genuine abandonment with friction. A cart abandoned because of a delivery cost surprise is recoverable by an ad that addresses it.

Cross-sell and repeat purchase in categories with natural replenishment cycles.

Audiences excluded from converters. Continuing to advertise to people who already bought is the most common waste in remarketing and among the easiest to fix.

Frequency

Diminishing returns arrive fast and turn negative. Beyond a modest number of impressions per person per week, additional exposure produces irritation rather than conversion, and it is invisible in performance data because irritated people simply never come back.

Cap it. The right number is lower than the default and lower than feels natural.

The reporting fix

Report remarketing separately from prospecting, always. A blended cost per acquisition across both is dominated by remarketing's flattering numbers and tells you nothing about whether you can acquire new customers profitably — which is the only question that determines whether the account can grow.