The short answer
Remarketing shows ads to people who have already visited your website or engaged with your content. Because they know who you are, the click usually costs less and converts better than cold traffic. The work that matters is splitting those visitors into segments worth different money, then capping how often each sees you.
What the work covers
- Tracking and audience setup across Google, Meta and Microsoft, checked not assumed
- Segments built by behaviour and value instead of one list of everyone
- Exclusions applied, starting with people who already bought
- Frequency caps set per segment and reviewed against performance
- Creative written per segment, because a cart abandoner and a blog reader need different ads
- Window lengths matched to the buying decision, plus server side tagging where the browser gets in the way
Segmentation and frequency, where almost every retargeting campaign goes wrong
Two settings decide whether retargeting makes money or quietly wastes it, and both usually sit on the default.
One audience of all site visitors is close to useless.
That list holds somebody who read a blog post for eight seconds, somebody who abandoned a $3,000 cart, three of your own staff, a competitor checking your pricing and every customer who already bought. Then you bid the same to reach all of them and show them the same ad. The cart abandoner is worth twenty times the blog reader, and the customer who bought yesterday is worth nothing, because you're paying to advertise a product they own.
The segments that actually earn.
Cart and checkout abandoners first, closest to the money and worth the highest bid and the most urgent creative. Then people who viewed a specific service or product page, retargeted with an ad about that thing rather than your brand in general. Then form starters who didn't submit, a small audience and often the best converting one you'll ever run.
Exclusions matter as much as inclusions.
Existing customers come out, unless you're deliberately running a cross sell or renewal audience, which is a different campaign with different creative. Converters come out the moment they convert, which needs the conversion event wired into the audience rather than someone remembering. Your own office IP comes out too.
Frequency capping is commercial, not a courtesy.
Impressions have sharply diminishing returns. The first few do the reminding. After that the same person keeps costing you money without becoming any likelier to buy, and the budget would have earned more shown to somebody new. There's a reputational cost too, measurable on Meta as negative feedback. We cap per user per week rather than per day, tighter on small audiences where the same hundred people absorb the whole budget, and use a burn pixel so ads stop the moment someone converts.
Window length depends on what you sell.
Emergency and impulse services want a short window, roughly three to seven days, because after that the job's been done by someone else. Considered retail and mid value services sit around fourteen to thirty. Property, finance and anything with a committee behind it runs sixty to a hundred and eighty days. Get it wrong either way and you stop too early or spend months following people who bought elsewhere in week two.
Then the privacy shift.
Third party cookies have been restricted for years in Safari and Firefox, iOS asks whether apps may track users, and tracking prevention keeps tightening. Some browsers truncate client side cookies to days. Audiences shrink, long windows quietly stop working, and small lists fall under platform minimums and never serve.
What survives is first party data. Server side tagging so measurement isn't at the mercy of a browser extension, customer lists uploaded and matched, consent mode implemented properly, and enhanced conversions where available. Email and SMS follow up stop being a nice extra and become the retargeting nobody can block.
How we set it up
We check the tracking
Plenty of accounts we inherit have audiences that stopped collecting months ago and nobody noticed.
We build the segments
Behaviour and value, with exclusions applied before a dollar goes out.
We write per segment
The ad a cart abandoner sees shouldn't be the ad a first time reader sees.
We cap it and watch it
Frequency, window length and creative fatigue reviewed monthly.
Who this suits, and who it doesn't
It suits any business with enough traffic to build a list, a considered purchase, or an eCommerce store where abandoned carts are a daily event.
It's a poor fit if traffic is too small to reach platform minimums, or if you'd be retargeting people who arrived on irrelevant searches. Fix the traffic first.
Why clients run retargeting here
Segments, not one list.
The audience structure is the work. Anyone can install a pixel and switch on a campaign that follows everybody.
One team across the platforms.
Google, Meta and Microsoft run from the same room, so one person sees the whole frequency picture instead of a single platform's slice.
No lock in contracts.
Month to month, and we'll cut the spend ourselves when an audience is saturated.
"Doesn't this just annoy people?"
It does when it's uncapped, unsegmented and still chasing people who bought a month ago. Capped and segmented, most people don't consciously notice it. The annoying version is also the expensive one, which makes this easy to agree on.