Retargeting & Remarketing
Retargeting works right up until it becomes the reason people avoid you
It is the easiest campaign to launch and the easiest to run badly. Almost every problem here is a restraint problem: the audience is too broad, the cap is too high, the buyers were never excluded, and the window outlasts anybody’s real buying cycle.
Almost every retargeting problem is a restraint problem. The audience is too broad, the cap is too high, the people who already bought were never excluded, and the window is longer than any real buying cycle.
The tactic is genuinely useful. Someone came, considered, and left without acting, and a well-timed reminder with something new to say is a reasonable thing to send them. The trouble is that the same mechanism, left unbounded, turns into pursuit.
This page is organised around limits rather than reach because limits are what almost every account we inherit is missing. The audience settings take an afternoon. Deciding when to stop is the actual work.
The limits we work to
What we will do with your audiences, and what we will not
We will
- Set a frequency cap on every campaign type that supports one, and record why that number was chosen
- Build and apply exclusion audiences for people who have already bought, and for accounts already in an active sales process
- Segment by what someone actually did, so a person who reached checkout is not treated like a person who bounced off the homepage
- Set membership and conversion windows against your real buying cycle rather than accepting whichever default the platform offers
- Run a holdout test where volume allows it, and report the result even when it undermines the case for the work
We will not
- Run a single all-visitors audience as the whole programme, because the message is then wrong for most of the people receiving it
- Show dynamic product advertising for an item the person has already purchased
- Upload or use customer data outside the consent it was collected under, or buy third-party lists
- Claim retargeting caused a conversion when the honest description is that it was present for one
- Keep scaling a tactic because its reported return is flattering when the underlying evidence has not been tested
How it turns
Four ways a useful tactic becomes a liability
- One audience holds everyone who ever loaded a page.
- A person who read a blog post for nine seconds and a person who abandoned a full basket are in the same bucket, receiving the same message. The result is an ad that is slightly wrong for almost everyone in it, and there is no creative fix for an audience definition problem.
- Nobody built the purchaser exclusion.
- Conversion audiences do not exclude themselves. Unless someone creates the list and applies it as an exclusion on every relevant campaign, you will keep spending to advertise a product to its new owner. Customers notice this more than any other advertising error, and they draw conclusions about the company from it.
- The window outlasts the decision.
- A ninety-day membership window on a product people buy within a week means most of the audience is made up of people who decided long ago, one way or the other. Long windows also inflate the reported effect, because they widen the net of conversions that can be associated with an impression.
- The audience is quietly shrinking and nothing is replacing it.
- Restrictions on cross-site cookies keep narrowing what behaviour-based audiences can hold and what can be reported about them. Accounts that depend entirely on that signal are on a slow decline, and the replacement, first-party data collected with proper consent, takes months of accumulation before it is useful.
Scope
What the work covers
- Full audit of existing audiences: what each one contains, how it was defined and whether anyone still knows why
- Segmentation by behaviour and depth, with a message that suits each segment rather than one message stretched over all of them
- Exclusion architecture: purchasers, existing customers, current opportunities, job applicants and anyone else who should not be pursued
- Frequency caps set per campaign, with the number recorded and revisited rather than set once
- Membership and conversion windows aligned to observed time-to-decision instead of platform defaults
- Creative sequencing so the second and third exposures say something new
- First-party audience building from consented customer records and owned engagement
- Suppression coordination with email so the same person is not receiving both at full volume
- Holdout testing where the audience is large enough to make the comparison fair
- Reporting that distinguishes assisted conversions from claimed causation
Questions
What people ask about following visitors around
Our retargeting shows the strongest return in the account. Why change it?
Because that number is measuring presence, not persuasion. Retargeting is shown to people who have already visited, which means it is systematically placed in front of the audience most likely to convert regardless of what it does.
The way to find out how much it is contributing is a holdout: withhold retargeting from a portion of the eligible audience and compare. Most accounts have never run one, and the ones that do are usually surprised in both directions.
How often is too often?
There is no single correct number, but uncapped is almost always wrong. Frequency capping in Google Ads limits impressions or views per person over a day, week or month, and it is not supported on every campaign type.
We set it against the length of the message and the length of the buying cycle. Something considered over three months tolerates a slower drumbeat; a same-week purchase does not need three weeks of reminders.
Why do people keep seeing ads for something they already bought?
Because nobody built the exclusion. Purchase and conversion audiences have to be created and then actively excluded from the campaigns, and that is a separate piece of work from creating the retargeting list in the first place.
It is worth fixing beyond the wasted spend. Advertising a product to the person who owns it is the clearest possible signal that the company is not paying attention.
What happens as cookies keep getting restricted?
Third-party signal keeps narrowing, so audiences built from browsing behaviour get smaller and less reliable, and some measurement associated with them stops being reportable at all.
The practical answer is to shift weight towards data you hold under your own consent: customer lists, engagement with your own content and app, and email. That work takes months to build, which is why it should start before the shortfall is urgent.
We are a small business with modest traffic. Is this relevant?
Partly. Below a certain volume, audience lists never reach the minimum sizes platforms require, and segmentation splits a small pool into several unusable ones.
In that situation the honest advice is usually one carefully bounded audience with a short window, a firm cap, and buyers excluded, or to put the money into the landing experience instead. We will tell you which of those applies.
Find out how much of your retargeting is actually doing work
We will look at what your audiences contain, what is excluded, how often people are seeing you, and whether the reported return survives being tested. Often the recommendation is to spend less here and more somewhere else.
If we don't deliver the work we agreed to deliver for reasons within our control, you don't pay for the undelivered work. Read our guarantee
Last updated · Reviewed by Zubair Afzal