Remarketing: How It Works and When to Stop
Every guide explains how to switch it on. Almost none explain the harder question — the point at which it stops adding anything and starts just following people around.
Remarketing is the campaign type that always looks like it is working.
The cost per acquisition is low. The conversion rate beats everything else in the account. Somebody suggests putting more budget behind it.
And they might be right. But the numbers look good partly because you are advertising to people who had already decided — and no dashboard separates the conversions you caused from the ones you simply witnessed.
Remarketing or retargeting?
Same thing. Google historically said remarketing, most other platforms said retargeting, and the industry now uses both interchangeably.
Pick one word and use it consistently in your reporting. That is the whole answer.
What you can actually build an audience from
The signal you choose defines the intent you are buying.
| Audience | Intent level | Best used for |
|---|---|---|
| Cart or checkout abandoners | Highest | A direct nudge — the offer, the reassurance, the missing detail |
| Product or pricing page visitors | High | Proof and objection handling |
| Any site visitor | Mixed | Broad reach, but dilute — segment it if you can |
| Video viewers (25%, 50%, 75%) | Rising with depth | Warming people who engaged but never visited |
| Social engagers | Low to medium | Cheap top-of-funnel reach, useful for scale |
| Customer lists | Highest, different job | Upsell, cross-sell, win-back — or exclusion |
A single "all visitors" audience is the default and the weakest option. Someone who bounced off your blog in nine seconds and someone who abandoned a full cart are not the same person, and showing them the same ad wastes both impressions.
Window length: match the buying cycle
The default window is whatever the platform suggested. That number has nothing to do with your business.
- Find out how long your buyers actually take. Time from first visit to purchase, from your own analytics or CRM. If nobody knows, that is the first thing to fix.
- Set the window around that period, with a little room either side. Short cycles want short windows so the audience stays warm.
- Segment by recency where volume allows. People from the last three days deserve a different message from people at day twenty-five.
- Escalate the offer as recency drops — a reminder early, a stronger reason to act later. Not the identical ad for a month.
- Cap the total window. Beyond your buying cycle you are paying for impressions with little realistic chance of converting.
- Revisit it quarterly. Buying cycles change with price, competition and season.
Exclusions matter more than inclusions
Most remarketing waste is not about who you targeted. It is about who you failed to remove.
- Recent converters. The ad that follows someone's purchase around for two weeks is the single most common complaint about this channel, and it is entirely self-inflicted.
- Existing customers, unless you are deliberately selling them something else — in which case build them a separate campaign with a different message.
- Job applicants and your own team. Careers page traffic in a sales remarketing audience is pure waste.
- Anyone past the window. Set it and enforce it.
- Mobile app placements, if you are running display. Accidental clicks in games are a well-known drain — see display advertising.
Frequency — and the point where it turns negative
There is no universal frequency cap worth quoting. What matters is the direction your own numbers move.
Watch for frequency climbing while conversions stay flat, on a stable audience and budget. That combination means you are showing the same message to the same people, and the ones who were going to act already have.
The incrementality question
This is the one that decides whether the budget is justified, and almost nobody runs it.
Remarketing targets people already inclined to buy. Some of them would have returned anyway. Last-click attribution credits the remarketing ad for all of them.
The honest test is a holdout. Exclude a random portion of your remarketing audience from seeing the ads, then compare conversion rates between the excluded group and the targeted group. The difference is what remarketing actually caused.
It usually finds that remarketing is genuinely additive — but by less than the dashboard claims. That is a useful thing to know before you scale the budget, and it is the same measurement discipline described in conversion tracking.
Platform notes
Meta. Audiences are built from Pixel and Conversions API events, so the quality of your remarketing depends entirely on whether tracking is set up properly — covered in Meta Pixel and Conversions API. Custom audiences from site behaviour, and lookalikes when you want to expand beyond them.
Google. Audiences come from tag or Analytics data and can be applied to search, display, YouTube and Shopping. Search remarketing — bidding higher for past visitors on the same queries — is underused and often the most efficient version of it.
LinkedIn. Website and engagement audiences work, but minimum audience sizes are large and the cost per impression is high. Worth it only where the deal value carries it. In practice, the pattern that works for B2B is reaching people on LinkedIn first, then staying in front of them on Meta at a fraction of the cost.
Frequently asked questions
What is the difference between remarketing and retargeting?
Nothing meaningful. Google historically called it remarketing and most other platforms called it retargeting, so the two words describe the same thing — showing ads to people who have already interacted with you. Use whichever your team prefers, but pick one so your reporting stays consistent.
How long should a remarketing window be?
Match it to your buying cycle, not to a default. Someone browsing a cheap product decides in days, so a short window keeps the audience relevant. A considered B2B purchase runs months, so a longer window makes sense. If you cannot say how long your buyers take to decide, that is the question to answer first.
Does remarketing actually work, or would those people have bought anyway?
That is the right question and most reporting hides it. Remarketing targets people already inclined to buy, so it takes credit for conversions that would have happened regardless. The only honest test is a holdout — exclude a portion of the audience and compare conversion rates between the two groups.
What should I exclude from remarketing audiences?
Existing customers, unless you are deliberately selling them something else. Recent converters, who do not need the ad that follows their purchase. Job applicants and your own staff. And anyone past your buying window, who is costing you impressions without any realistic chance of converting.
How high can remarketing frequency go before it hurts?
Watch your own numbers rather than a universal figure. When frequency climbs while conversions stay flat, you are paying to annoy people who already decided not to buy. Rising frequency alongside a rising cost per result on a stable audience is the signal to cap it or shorten the window.
We test whether it is actually incremental
Segmented audiences, disciplined exclusions, and holdout tests that separate the conversions you caused from the ones you witnessed.