Remarketing that does not burn budget: what to split and what to exclude

Remarketing is the best-ROAS campaign in nearly every account, and the most misleading. It talks to people who already know the brand, already visited the site, already showed interest — and naturally converts more.

The question almost nobody asks: how many of those sales would have happened without the ad?

The problem of undeserved credit

Someone visited the site yesterday, is deciding, and would come back today regardless. If along the way they see a remarketing ad and click it, the platform records the sale as the ad's doing.

You paid for a click that replaced a direct visit which would have been free.

That does not invalidate remarketing — it invalidates reading its ROAS as if it were incremental. Remarketing ROAS measures audience quality, not campaign contribution.

The four lists that work

Generic remarketing to "everyone who visited the site" wastes budget on people who read the blog and never considered buying. Split by intent:

Each deserves a different message, because each stopped in a different place. Someone who abandoned the form needs a nudge; someone who read the blog needs a reason.

The video-view list deserves a note: watching 75% of a two-minute video is an interest signal comparable to visiting the pricing page, and it is one of the cheapest lists to build.

What to always exclude

Exclusion is what separates remarketing from waste:

Anyone who already bought, except where genuine repurchase happens in the period. Advertising to someone who just bought annoys them and pays for nothing.

Anyone in active negotiation. Sales is already talking to that person. The ad adds nothing and sometimes gets in the way, showing an offer different from the one being negotiated.

Anyone disqualified. Outside the region, outside the profile, no budget. That list has to come from the CRM, and it is the exclusion that saves the most.

Anyone who already converted on the event you are optimizing for. It looks obvious and it is the most common error in hastily built accounts.

The first three require the CRM feeding the platform — custom audiences kept updated with who bought, who is in negotiation and who was discarded. Without that, you are advertising to your own customer base.

Frequency: the tipping point

Remarketing has a small audience, so frequency rises fast. And here it has an effect that does not exist in prospecting: it irritates.

The same person seeing the same ad eight times in three days does not become more likely to buy. They become annoyed with the brand.

Use a frequency cap, vary the pieces within the list, and close the window. Someone who has not converted in 30 days rarely converts on day 45 through persistence — and keeps costing money and goodwill.

How to measure what it actually adds

Two ways, one cheap and one good.

The cheap one: watch the total

Pause remarketing for two weeks and watch total company sales, not the campaign's.

If the total falls in proportion to what remarketing was claiming, it was bringing new sales. If the total barely moves, much of those sales were being captured, not created.

It is crude, it is frightening, and it gives the most honest answer most accounts will ever get.

The good one: a holdout test

Hold back a slice of the audience from seeing the ad and compare the conversion rate of both groups. The difference is the real increment.

It needs volume, discipline and a platform that supports the design. Worth it when remarketing is a large slice of the budget.

The healthy proportion

10 to 20% of the budget covers remarketing well in most accounts.

Above 30%, it is almost certain you are paying again for people who were coming anyway — and, worse, not investing in prospecting, which is what feeds next month's remarketing list.

Remarketing without prospecting is a shrinking list. ROAS looks beautiful for two months and the account dies in the third.

The summary

Split lists by intent, not by "visited the site". Exclude customers, active negotiations and disqualified contacts — using the CRM. Cap frequency and close the window. Keep it between 10 and 20% of budget. And read remarketing ROAS knowing it measures the audience, not the merit.

For splitting budget across each campaign's function, see splitting budget across channels.