What is Remarketing?

Remarketing advertises to people who have already interacted with a business. The audiences are built from observed behaviour, and the sources fall into a few groups.

Site and app activity. Visited any page, viewed a particular product, searched within the site, added to a cart, began a checkout, reached a stage in a signup. Captured by a pixel or, more durably, by server-side events.

Known customers and contacts. A list the business holds, matched to platform accounts.

Platform engagement. People who watched a share of a video, opened a lead form, saved a post, interacted with a profile. This data lives on the platform and is available only there.

Each carries a different strength of signal. Somebody who abandoned a cart yesterday is a different proposition from somebody who read a blog post four months ago, and treating them as one audience is where most remarketing goes wrong.

Every remarketing audience has a membership window: how long a person stays in the pool after the qualifying behaviour.

It is set once, usually at a default, and it decides more about campaign performance than the bidding does.

The mechanism is frequency. Frequency is impressions divided by reach, and a long window produces a pool dominated by people whose interest has faded while a steady budget keeps spending against it. The result is high repetition against exactly the people least likely to respond, which produces falling response, rising negative feedback and a climbing cost per result — all of which look like fatigue and are actually a window problem.

The right length is the purchase cycle. A considered B2B purchase with a two-month cycle justifies a long window. A low-price impulse product does not, and a thirty-day window on it means four weeks of advertising to people who decided in the first two days.

Better than one long window is several short ones with different messages: the first three days say one thing, the following two weeks say another, and beyond that the person leaves or moves to a much cheaper, lower-frequency campaign.

A single audience of everyone who visited is the least useful version of this.

Depth of interaction ranks people usefully and costs nothing to implement. A cart abandoner is closer to buying than a product viewer, who is closer than a homepage visitor, who is closer than somebody who read an article. Each deserves a different budget, a different message and a different window.

Recency ranks them again. The same behaviour two days ago and two months ago are different signals.

And exclusions make the segments real: a cart abandoner audience that still contains product viewers is not a cart abandoner audience, and the campaign aimed at it will not behave as intended.

Remarketing campaigns report the lowest cost per acquisition in almost every account, and that figure is both accurate and misleading.

It is accurate because the conversions did happen and the spend was real. It is misleading because a substantial share of those people were going to convert anyway. Advertising to somebody who has a full cart and an intention to buy will be credited with the purchase whether or not it changed anything.

The gap between reported efficiency and actual contribution cannot be closed by better attribution, because attribution measures credit rather than causation. It can be measured directly, by holding out a portion of the audience and comparing, and accounts that run that test are usually surprised in the same direction.

None of this means remarketing does not work. It means the reported number overstates it, that scaling budget on the strength of that number has a ceiling, and that the honest comparison against prospecting requires a test rather than a report.

It cannot grow an audience. Every remarketing pool is downstream of something else that brought people in, so a business that funds remarketing at the expense of prospecting is spending against a pool that shrinks.

It cannot fix a weak offer. People who visited and left had a reason, and showing them the same page again addresses none of it.

And it saturates quickly. The pools are small by construction, which is why frequency is the recurring problem and why creative rotation matters more here than in prospecting.

Set the window to the purchase cycle, and split it into stages rather than running one long one.

Segment by depth — cart, product, category, general — and give each its own message.

Exclude converters, everywhere except a deliberate repeat-purchase campaign.

Cap the budget to the pool size rather than to ambition, because a pool cannot absorb more than a certain amount before it is only buying repetition.

Rotate creative faster than in prospecting, on a cadence derived from how quickly frequency climbs against the pool.

Test the increment at least once, so that the account knows the size of the gap between what is reported and what is caused.

The same activity behaves differently depending on where it runs, and the differences change what is worth building.

Search. Lists are applied to keyword campaigns, either to adjust bidding for people who have visited or to allow broader keywords only for them. The second is the more interesting use: terms too expensive or too vague to buy from strangers can be viable against people who already know the business.

Social feeds. The classic dynamic version, where a catalogue serves the product a person looked at. Effective and prone to every problem above, because the audiences are small and the delivery is continuous.

Display. The cheapest inventory and the widest reach, which makes it the surface where over-exposure is most likely and where the badly-run version of this activity got its reputation.

Video. Underused. A person who watched most of a video has self-selected in a way a page view does not match, and the resulting pool is warmer than its size suggests.

Email and owned channels. Free of auction costs entirely, and where the strict definition of the word would place it. An audience worth advertising to is usually an audience worth emailing first.

That last point is worth dwelling on. The cheapest remarketing is the kind that does not go through an auction, and an account paying to reach people whose addresses it already holds has an ordering problem rather than a bidding one.

Nothing separates these two words in practice, and that is the honest answer. Both describe advertising to people who have already interacted with a business, both are used for the same campaigns, and the platforms use them inconsistently themselves — Google Ads has long said remarketing, most social platforms say retargeting, and the same ad set would be called either depending on who is describing it. Where a distinction is claimed, it is usually that remarketing covers owned channels such as email to a known list while retargeting covers paid media to an anonymous cookie or device. That split is coherent, it is not universally observed, and an advertiser who assumes the other party is using it will misunderstand them about half the time. The useful discipline is not to pick a side but to say what is actually being described: which audience, built from which behaviour, over which window, reached on which surface.