In short: Retargeting shows ads to people who already know the company, which makes it cheap and makes its attributed results unreliable. Exclude people who converted, cap frequency, shorten windows, test with a holdout, and remember that email is retargeting you own.
Key points
- Retargeting is credited with many conversions that would have happened anyway because it stands closest to checkout.
- Incrementality studies consistently find its true contribution is a fraction of its attributed one.
- Showing ads to people who already bought is the most common and avoidable waste.
- Most incremental effect comes in the first days after a visit, not over ninety-day windows.
- Hold out a random share of the audience; if it converts nearly as well, move the budget.
Retargeting, also called remarketing, is advertising shown to people who have already visited a website, used an app, or otherwise identified themselves to a company. It is cheap, it reports extraordinary results, and it is the channel most likely to be taking credit for customers who were coming anyway.
How it works
A visitor arrives. A pixel or server-side event records the visit and adds the visitor to an audience. Ad platforms then show that audience the company's ads as they move around the web and social platforms. Audiences can be segmented: everyone who visited, people who viewed pricing, people who started an application and stopped, existing customers to be excluded or upsold.
As third-party cookies have declined, retargeting has shifted toward audiences built from first-party data, email lists matched to platforms, logged-in users, and platform-internal signals, and toward the walled gardens where the platform can recognize its own users without cookies.
What it is good for
Bringing back people who were interrupted: an application abandoned at the identity step, a cart left, a trial that stalled. Reminding people who compared options and left. Reaching existing customers with something new. Sequencing a message: a visitor who read an explainer sees proof next, then an offer.
For considered purchases, which describes most financial and business products, some of that is valuable. People do return, and a well-timed reminder with the right message does move some of them.
Why the reported results are inflated
Retargeting reaches people who already know the company and were already likely to convert. Under any attribution model that counts a view or a click before conversion, retargeting is credited with a large share of conversions that would have happened without it.
The pattern is consistent across incrementality studies: retargeting's true contribution is a fraction of its attributed one, often a small fraction, particularly for brand-name searchers and recent visitors who would have returned regardless. The channel looks like the best performer in the account because it stands closest to the checkout.
This is not an argument against retargeting. It is an argument against believing its dashboard.
How to run it without paying twice
Exclude people who already converted, unless the message is specifically for them. Paying to show an ad to someone who already bought is the most common and most avoidable waste.
Cap frequency. Following a person around the internet with the same ad for a month annoys them and does not work better than three or four exposures.
Segment by intent and message accordingly. An abandoned application needs a different message from a first visit.
Set a short window. Most of the incremental effect is in the first days after a visit. Audiences that stretch to ninety days are mostly paying for people who have moved on or will return anyway.
Test it. Hold out a random share of the audience and compare conversion rates. If the held-out group converts nearly as well, the channel is observing, not causing, and its budget belongs elsewhere.
For restricted categories
Retargeting depends on the platforms that carry the category. Where the large platforms will not run the ads, retargeting through them is not available, and the company relies on programmatic retargeting against its own audiences and on email, which is retargeting the company owns. That is a reasonable place to be: email reaches the same people at a fraction of the cost and does not need anyone's approval.
lowob takeaway: Retargeting reaches people who already know you, which makes it cheap and makes its attributed results unreliable. Exclude converters, cap frequency, shorten windows, test with a holdout, and remember that email is retargeting you own.
Common questions
What is retargeting?
Advertising shown to people who have already visited a website, used an app or otherwise identified themselves to a company. It is also called remarketing.
Why does retargeting look so effective in reports?
It reaches people already likely to convert, so attribution models credit it with conversions that would have happened anyway. Its true contribution is a fraction of its attributed one.
How do I test whether retargeting actually works?
Hold out a random share of the audience and compare conversion rates. If the held-out group converts nearly as well, the channel is observing, not causing.
What can restricted-category companies use instead of platform retargeting?
Programmatic retargeting against their own audiences and email, which reaches the same people at a fraction of the cost without anyone's approval.