A retargeting strategy that works splits warm audiences by how close they are to buying and how recently they showed it, gives each group a different message, and caps how often they see it. Then it judges the results skeptically, because retargeting usually reports the best ROAS in an account while causing less of it than the number suggests. Cut the budget when click-only results and holdout tests stop justifying it.
Why retargeting looks better than it is
Retargeting targets people who already visited your site, and many of them were coming back anyway. Three things inflate the numbers:
- Buyers who would have converted anyway. Someone who added to cart yesterday is already likely to buy. Show them an ad and the platform claims the sale, even if the ad changed nothing. Retargeting selects the audience most likely to convert, then takes credit for converting it.
- View-through credit. Meta’s default attribution setting counts a purchase within one day of someone seeing an ad, with no click needed. Warm audiences see ads constantly, so a large share of retargeting “conversions” can be impressions the buyer scrolled past on the way to your site.
- Audience overlap. The same cart abandoner sits in your Meta, Google and TikTok retargeting pools and in your abandoned-cart email flow. When they buy, every one of them may claim the order. Added together, platform-reported retargeting revenue can exceed the actual revenue from that group.
A hypothetical to make it concrete: 1,000 cart abandoners, and without any ads 150 of them come back and buy within a week. You retarget them and 170 buy. The platform can report up to 170 conversions; the ads caused 20. Judged on the reported number, retargeting looks more than eight times better than it is.
None of this means retargeting doesn’t work. It means the dashboard can’t tell you how well.
Segmenting by intent and recency
Two things decide what a warm visitor is worth: what they did (intent) and how long ago (recency). Build segments from both, and make recency windows mutually exclusive by excluding each shorter window from the longer one, so every person sits in one segment at a time.
| Segment | Ecommerce signal | B2B signal | Starting window |
|---|---|---|---|
| High intent | Added to cart, started checkout | Visited pricing or demo page, started a form | 1-7 days |
| Mid intent | Viewed product pages, several page views | Read case studies, integration or product pages | 8-30 days |
| Low intent | Single visit, blog readers, video viewers, social engagers | Blog readers, webinar registrants, video viewers | 31-90 days |
| Customers | Past purchasers | Customers and open opportunities | Separate program, usually excluded |
The windows are starting points, not rules. Set them from your own data: check how many days typically pass between first visit and purchase or demo request, and make the high-intent window cover most of the quick converters.
Two constraints limit how finely you can slice. First, each segment must be big enough to deliver. Platforms need a minimum matched audience, and a tiny ad set either won’t spend or keeps hitting the same few people. Second, every extra ad set splits the platform’s learning. If your traffic is modest, three segments (high, mid, low) are plenty. Merge before you split.
Customers belong in retention, not retargeting. Cross-sell and replenishment ads to past buyers can make sense, but they need their own budget and their own measurement so they don’t inflate the retargeting line.
Messages for each segment
Each segment has a different reason it didn’t convert, so write to the objection, not to the product.
- Cart and checkout abandoners: remove friction. Shipping cost and speed, returns policy, reviews on the exact product, payment options. Don’t open with a discount; people quickly learn that abandoning a cart earns a code.
- Product and category viewers: dynamic product ads showing what they viewed plus close alternatives, with social proof and the one differentiator that matters in your category.
- Low-intent visitors and engagers: a reason to come back, such as a bestseller, a new collection, a useful guide, or a founder or customer video. Asking a blog reader to “buy now” wastes the impression.
- B2B pricing and demo-page visitors: proof and risk reduction. A customer result in their industry, a short product walkthrough, security or implementation answers, or a lighter next step than a full demo.
- B2B content readers: the next piece of content in the journey, not “book a demo.”
Warm audiences are small and see ads often, so creative wears out faster than in prospecting. Refresh retargeting creative on a set rhythm and keep at least two or three concepts running per segment.
Frequency caps and exclusions
Frequency is where retargeting turns from helpful to irritating. Where a platform lets you set a cap, as Google display campaigns and reach-optimized campaigns on Meta do, set one. On conversion-optimized campaigns without a hard cap, frequency is driven by budget relative to audience size, so budget is your control.
A hypothetical check: a 20,000-person pool, a $15 CPM and a $300 weekly budget buy 20,000 impressions a week. If the platform reaches 60% of the pool, that’s about 1.7 impressions per reached person per week. Double the budget without growing the pool and frequency roughly doubles, while results rarely do.
My starting points, not benchmarks: a higher frequency is tolerable for a few days after a high-intent action, then it should drop sharply, and long low-intent windows should see an ad once or twice a week at most. Check the frequency column weekly. When frequency and cost per result rise together, the pool is saturated.
Exclusions do as much work as targeting:
- Recent purchasers, for at least the length of your typical repurchase cycle
- Customers and open opportunities from the CRM, uploaded as a customer list and refreshed automatically
- Converted leads, so people who just booked a demo stop seeing “book a demo”
- Careers page visitors, logged-in app users and support page visitors
- Shorter recency windows excluded from longer ones
- Cart abandoners still inside your abandoned-cart email or SMS sequence, or at least a delay before paid ads start
- Your own team and agency, by email list or IP where the platform allows
Retargeting for B2B vs ecommerce
The mechanics are the same; the economics are not.
| Ecommerce | B2B | |
|---|---|---|
| Pool size | Large, refilled daily by prospecting | Small, often near minimum-audience limits |
| Time to convert | Hours to weeks | Weeks to months |
| Main format | Dynamic product ads | Proof, content and demo offers |
| Main platforms | Meta, Google dynamic remarketing and PMax, TikTok | LinkedIn, Google, and Meta for SMB or founder-led buyers |
| Competes with | Email and SMS flows | Sales follow-up and nurture email |
| Success metric | Incremental contribution margin | Qualified pipeline |
In ecommerce, the biggest risk is paying for sales your email and SMS flows would have recovered for free. In B2B, the risk is optimizing for cheap form fills from people who were already talking to sales. Judge B2B retargeting on qualified pipeline in the CRM, not on platform-reported leads. Company-level targeting on LinkedIn, built from a target account list, can also do what small website pools can’t: keep your proof in front of the buying committee, not just the one person who visited.
Reading retargeting results honestly
Before you trust a retargeting number, run four checks:
- Compare click-only and default attribution. In Meta Ads Manager, use the compare attribution settings option to see click-only results next to the default. A large gap means much of the credit comes from views.
- Check new vs returning buyers. If most retargeting conversions come from existing customers, you’re paying to reach people your owned channels already reach.
- Add up the overlap. Sum what Meta, Google and your email platform each attribute to the same audience over a month, then compare with the actual orders from those people. The difference is double counting.
- Watch blended results when retargeting spend changes. If retargeting spend moves and total revenue or pipeline barely does, the ads were collecting credit, not creating sales.
These checks point in a direction; only a holdout gives an answer. On/off tests and platform lift studies both suit retargeting, and the incrementality testing guide covers how to design one.
When to cut retargeting budget
Cut, or at least shrink, retargeting when any of these hold:
- Click-only ROAS or CPA is worse than break-even, even if the default view looks healthy.
- Frequency keeps climbing and cost per result climbs with it.
- A holdout shows little or no lift.
- Retargeting’s share of paid spend grows while prospecting shrinks. Retargeting only harvests what prospecting plants, so starving prospecting eventually empties the pool.
- Broad automated campaigns already reach warm audiences. Meta’s Advantage+ campaigns and Google’s Performance Max both serve returning visitors, so a separate retargeting campaign may be competing with them for the same people.
Cut in order: long, low-intent windows first, then mid-intent, and keep short high-intent windows longest. Step down rather than switching off, and hold each level for two to four weeks while you watch blended revenue or pipeline. If blended results don’t move, keep the savings and move them into prospecting. The Meta ads account structure guide shows where retargeting fits beside prospecting at different budget levels.
Keep or grow retargeting when tests show real lift, when the sales cycle is long and the product is expensive, or when a specific segment, like checkout abandoners on a high-ticket product, keeps earning its place on click-only numbers. Setting up and pruning these layers across platforms is a regular part of my performance marketing work.
Get it built
If retargeting reports your best ROAS and you’re not sure it’s real, I’ll audit the segments, exclusions and overlap and show what it actually adds. The Growth Audit is $1,500 fixed and credited if we continue. See pricing or get in touch.