The typical retargeting setup is one audience — "all site visitors, 30 days" — seeing one ad. It underperforms for a predictable reason: someone who bounced off your homepage in nine seconds and someone who read three articles and started your contact form are not the same person, and showing them the same message serves neither.
For a considered B2B purchase, the majority of first-time visitors won't convert. That's not a failure of the page; it's the nature of buying something expensive. Retargeting exists to stay present through a decision that takes weeks.
Segment by what they did
Behaviour tells you where someone is. Build audiences accordingly:
| Segment | What it means | What to show |
|---|---|---|
| Bounced under ~15s | Wrong fit, or wrong page | Usually exclude — don't pay to chase them |
| Read one service page | Aware, early | Proof: a case study with real numbers |
| Read multiple articles | Actively researching | Something substantial — a guide, not a sales call |
| Visited pricing | Evaluating seriously | Address the likely objection directly |
| Started a form, abandoned | High intent, hit friction | Remove the friction; offer a lower-commitment path |
| Enquired, went quiet | Stalled internally | A specific, low-pressure reason to re-engage |
| Became a customer | Done | Exclude. Always. |
That last row matters more than it sounds. Continuing to serve acquisition ads to existing customers wastes budget and looks careless to the people who already trusted you.
Exclude the sub-15-second bounces too. They're the largest segment and the least likely to convert. Paying to re-reach them is where retargeting budgets quietly disappear.
What to actually show
The common failure is showing the same "Book a call" ad to everyone at every stage. Match the message to where they are.
Early — proof over persuasion. For someone who read one page, the question is "are these people credible?" Answer with evidence: a case study with real, specific numbers. Vague claims do nothing here; a concrete outcome does.
Middle — depth. Someone reading multiple articles is educating themselves. Give them more of what they came for. An offer of substantial material converts better than a sales push, and it keeps you in the consideration set.
Late — remove friction. Form abandoners hit something: too many fields, an unwanted phone call, unclear pricing. Offer the lower-commitment alternative — a message instead of a call, a scoped quote instead of a discovery meeting.
Stalled — a reason, not a reminder. "Still interested?" is weak. A relevant new case study, or a specific answer to the objection that likely stopped them, is not.
Frequency and duration
Two settings that cause most retargeting resentment.
Cap frequency. Someone seeing your ad several times a day for a month doesn't become more likely to buy; they become annoyed and associate your brand with being followed. A few impressions per week is plenty for a considered purchase.
Match window to your sales cycle. If deals take two months, a 30-day window drops people mid-decision. If your cycle is a week, a 90-day window pays to reach people who already chose someone else. Look at your actual CRM data for time-from-first-touch-to-close and set the window from that.
Rotate creative. The same image for six weeks stops being seen. Refresh on a schedule.
The part that usually matters more
Before spending more on retargeting, look at what people land on. Conversion work on your own site frequently returns more than another ad campaign, because it improves every channel at once rather than one.
Where the returns usually are, in order:
Proof with real numbers. The most common gap on agency and services sites. Case studies with [X] placeholders, or vague claims like "significantly improved performance," fail exactly the visitor you most want. A prospect evaluating an unfamiliar vendor is looking for evidence you've done this before and it worked. Specific numbers do that; adjectives don't.
A clear single action per page. Pages offering a call, a quote, a newsletter, and a download convert worse than pages offering one thing.
Reduced form friction. Every field costs completions. Ask for the minimum that lets you respond usefully, and collect the rest in the conversation.
Speed on mobile. You paid for that visitor. A slow page loses people you're being billed for.
A messaging option. For international B2B — particularly outside North America and Western Europe — a WhatsApp button often outperforms a form. Lower commitment than a call, faster than email, and the conversation starts immediately.
Visible credibility. Real client names where you're permitted to use them, genuine reviews, a real person's name and face behind the business. For an offshore or unfamiliar vendor this carries disproportionate weight.
Measuring it honestly
Retargeting attribution flatters itself. These are people who already knew you; some would have returned regardless. Last-click will happily credit retargeting for conversions it merely observed.
More honest approaches:
- Hold out a percentage of your retargeting audience and compare conversion rates. This is the only genuinely clean measurement, and it's worth the small cost.
- Watch total qualified pipeline, not retargeting-attributed conversions.
- Check assisted conversions rather than last-click alone.
If pausing retargeting for two weeks doesn't visibly affect qualified pipeline, you've learned something valuable and cheap.
A sensible setup
- Exclude converters and sub-15-second bounces
- Three segments: read a service page, read multiple articles, abandoned a form
- One message per segment, matched to stage
- Frequency capped, window matched to your real sales cycle
- Fix the proof and friction problems on your site first
- Hold out a slice to measure whether any of it works
Most accounts are better served by step 5 than by steps 1 through 4.
Want the funnel and the site working together?
We build the landing pages, tracking, and retargeting structure as one system — and we'll tell you when the site is the problem rather than the ads. Talk to us about digital marketing.