Bottom of funnel
Retargeting, with the incrementality question answered.
The reported return on a retargeting campaign is almost always excellent, because the audience was already the most likely group to buy before the ad ever ran. A withheld comparison group is the only way to find out whether the campaign itself did anything.
This page covers re-engaging people who have already visited a site, started a form, or interacted with a page in some other way. Reaching people who have not encountered the business before is a separate discipline with a different audience and a different economics, covered as paid social prospecting. A warm audience and a cold one respond to different creative, different frequency, and different measurement standards, and treating both the same way hides how differently each one actually performs.
Show an ad only to people who already put something in a cart, or who already looked at pricing twice this week, and the campaign will look like a spectacular performer — not because the ad is unusually persuasive, but because the audience it reached was already, on its own, more likely to buy than anyone else on the internet. That is retargeting's advantage and its measurement problem at the same time. Audience selection alone accounts for most of the apparent return, and a standard platform report has no way to separate that from anything the ad itself actually caused.
There is a direct consequence of this that most vendors avoid saying plainly: without a comparison group that never saw the campaign, a retargeting report cannot be wrong. Every number it produces will look good, whether the ads changed a single outcome or changed nothing at all, because the report is measuring an audience that was already inclined to convert. Frequency-capped, well-sequenced re-engagement clearly does convert some genuine fence-sitters, and that is exactly why the one test that can tell the difference is worth insisting on.
How the work differs
- Audience definition
- Specific qualifying actions — cart abandonment, repeat pricing-page visits, an expired trial — set individually for each campaign.
- Frequency capping
- Impression limits set per person across the entire retargeting program, tracked at the account level.
- Sequencing and creative rotation
- Creative rotated on a schedule tied to how long someone has been in an audience, so week three shows a different message than week one.
- Exclusion logic
- Anyone who already converted removed from the qualifying audience within a set window, so budget stops chasing someone who already bought.
- Holdout comparison
- A comparable segment of the qualifying audience withheld from the campaign entirely for a defined period, with conversion rates compared between the two groups afterward.
- Incremental reporting
- The platform's own reported conversions logged separately from the holdout-measured lift, with budget decisions made on the lift number.
Why the reported return is almost always good
Every retargeting audience is, by definition, a group that already showed real intent — that is what qualifies someone for it in the first place. A platform's conversion count for that audience will therefore always look strong relative to a cold audience, because the two groups start from completely different baseline odds of converting, independent of anything the campaign itself contributed. Most retargeting reports quietly present that baseline difference as the campaign's own achievement.
This distortion is structural, shared by every platform and every vendor measuring a pre-qualified audience this way, and it comes entirely from how the audience itself was selected. The only way to see past it is to ask what that same audience would have done anyway, which requires deliberately withholding the campaign from a comparable slice of it and watching what happens.
How it is measured
The number reported is incremental lift: the gap between what the audience that saw the campaign did and what the comparable holdout group did on its own. Budget decisions run off that lift number inside the attribution layer, on the same terms as every other channel's incremental result.
What this channel cannot claim credit for: a purchase from someone who had already decided and was checking out at the moment the ad happened to appear. The holdout exists specifically to keep that share of "conversions" out of the number retargeting gets judged on.
What we need from you
Whoever manages the ad platform needs room to withhold a comparison segment from every retargeting campaign at setup, since carving out the holdout after a campaign is already running produces a comparison that is no longer clean. Access to the site or CRM events that define who qualifies for an audience in the first place — cart activity, pricing-page visits, trial expiration. And tolerance for a portion of an obviously interested audience seeing nothing for the length of the test, which is the part of this most teams find hardest to accept.
Who this is for
Built for operators with enough returning-visitor or cart-abandonment volume to support a real comparison group, part of the bottom-of-funnel lineup, where the current retargeting report already looks excellent and nobody has checked whether it would look the same with no campaign running at all. Self-storage operators see this constantly with unit-size and pricing-page visitors who leave without booking — a highly retargetable audience that also converts at a meaningfully elevated rate on its own, which is exactly the pattern a holdout is built to separate out.
Frequency here is coordinated with email and lifecycle marketing so the same person is not hit by ads and messages on the same day for the same offer, and the audience logic pairs closely with conversion rate optimization, since a retargeting-driven return visit is only worth the spend if the page it lands on actually converts. Client results show what a completed holdout comparison and its lift number look like once the test finishes.
Large enough that a withheld comparison group still produces a statistically meaningful number of conversions on its own, which depends entirely on the site's traffic and baseline conversion rate. A high-traffic e-commerce catalog can run a meaningful holdout inside a single week; a lower-volume B2B site with a long sales cycle may need to run the comparison over a full quarter to gather enough conversions on each side to trust the result.
Long enough to still be relevant to the decision, and not a day longer. Someone who abandoned a cart yesterday is a different prospect than someone who looked once three months ago and has likely already decided elsewhere, and showing both the same ad at the same frequency wastes budget on the second group while probably under-serving the first. Audience duration is set per product and sales cycle rather than a single default window applied everywhere.
A comparable segment of the qualifying audience is withheld from the campaign entirely for a defined period, and the conversion rate of that group is compared against the group that saw the ads. Any gap between the two is the campaign's actual, defensible effect; anything the platform reports beyond that gap is describing an audience that was likely to convert with or without the ad.
Past a certain point, yes — the same person seeing the same ad a dozen times a week tends to read as intrusive rather than persuasive, and can measurably suppress engagement instead of driving it. We cap impressions per person across the full program, not per individual ad set, since someone can hit an uncomfortable frequency without any single campaign appearing to be the culprit.
The audience-building mechanics shift toward first-party data — a business's own site visitors, cart activity, and CRM records — rather than a cross-site identifier supplied by a browser. The core measurement question does not change at all: a holdout comparison works the same way regardless of which technical method built the audience being tested, which is part of why we built the measurement standard around it rather than around any specific tracking method.
Find out what your retargeting is actually doing.
A holdout comparison on one active retargeting campaign, showing the real conversion gap between the audience that saw it and a matched group that did not.