A man buys a pair of running shoes on a Tuesday. On Wednesday, the same shoes appear in his social feed. On Thursday, they follow him onto a news site. By the following week he's seen them in his email sidebar, on a recipe blog and in the middle of a video he was trying to watch, all for shoes already sitting in his hallway. He doesn't think "what a clever brand". He thinks "this company is a bit creepy", and he mutes them.
Retargeting is one of the most effective tools in paid advertising and one of the most resented. It reaches people who have already shown interest, so its reported numbers usually look spectacular. But it's also the format most likely to feel intrusive when it's done carelessly: too often, for too long, after the purchase, with the same ad. And those spectacular numbers often hide a less flattering truth, because many of the people retargeted would have bought anyway. Good retargeting is a reminder, not a shadow. Get the windows, frequency and exclusions right, and it becomes both more welcome and more honest about what it's actually achieving.
This guide focuses on the settings that decide whether retargeting helps or annoys. Audience setup and platform basics are covered in retargeting done right.
Why retargeting annoys, and why it looks better than it is
Most annoying retargeting comes from five habits. Too frequent: the same person sees the same ad several times a day. Too long: ads keep appearing weeks after the person moved on. After the purchase: buyers are chased for something they already own. One message on repeat: the same creative, unchanged, again and again. No relevance: someone who read one blog post gets hard-sell product ads as if they'd abandoned a basket.
Meanwhile, retargeting reports usually look excellent. That's because the audience is already warm. Someone who added an item to their basket yesterday is far more likely to buy than a stranger, ads or no ads. When they do buy, the retargeting ad they saw on the way often receives the credit. The result is a format that can show very high return on ad spend while adding much less than it appears to, a measurement problem explored in why attribution is getting harder. Fixing the annoyance and fixing the measurement turn out to be closely linked: the same settings that make retargeting less irritating also stop it claiming credit it hasn't earned.
How long should a retargeting window be?
A retargeting window, or membership duration, is how long someone stays in an audience after an action. Set it to match how close that action was to buying and how long your customers usually take to decide: days for an abandoned basket, weeks for a category browser, months for a considered B2B purchase.
Membership isn't the same as delivery. Someone in a 30-day audience isn't shown ads for 30 days; they're simply eligible. Whether they see anything, and how often, depends on budget, bids, placements, competition and any frequency controls. So the window sets the outer limit of follow-up, while frequency and budget decide how intense it feels inside that limit.
| Visitor action | Window | Message |
|---|---|---|
| Abandoned basket or checkout | 1 to 7 days | Reminder, reassurance, help |
| Viewed a specific product | 3 to 14 days | The product, reviews, key benefit |
| Browsed a category | 14 to 30 days | Best-sellers, buying guides |
| Read content only | 30 to 90 days | More helpful content, not hard sell |
| B2B: viewed pricing or demo page | 30 to 90 days | Proof, case studies, objections |
| Past purchaser | Based on the repurchase cycle | Replenishment or genuine cross-sell |
Three factors should move these numbers. Price and consideration: expensive or considered purchases, like furniture, travel or B2B software, justify longer windows than everyday, low-cost products. Your own data: look at how many days typically pass between a first visit and a purchase, and set windows around that rather than a platform default. Repeat-purchase timing: for consumables such as coffee, skincare or pet food, the useful window for a past buyer opens shortly before they're likely to run out, not the day after they bought.
Long-consideration B2B purchases are the main case for windows of 60 to 90 days, because several people research over months; keep frequency low and the content useful across that period. Shorter windows are better for impulse and low-price products, flash sales and holiday offers with fixed end dates, and for any audience where click-through and conversion rates fall sharply after the first week. Longer windows don't automatically produce better results; they mostly add older, colder visitors.
Platforms set very different ceilings. Meta's Marketing API documentation lists a retention range of 1 to 180 days for website custom audiences, while Google Ads allows a membership duration of up to 540 days, with a 30-day default. Those are limits, not recommendations, and the gap between them is a good reason not to let defaults decide your windows. Basket abandoners also deserve a check on the experience itself, since many abandon for fixable reasons, as covered in optimising your checkout flow.
Landingi
Drag-and-drop landing page builder with reusable Smart Sections, A/B testing and 400+ templates.
Best for: Marketers and agencies building pixel-perfect landing pages without code.
How often should retargeting ads show?
There's no universal "right" frequency, but most retargeting works best with fewer impressions than advertisers expect. Frequency is the number of times one person sees your ads in a period. A daily limit stops bursts, such as the same ad five times in an evening; a weekly limit controls the total pressure someone feels. Higher-intent audiences, such as basket abandoners, can take more contact early on, while lower-intent audiences, such as blog readers, need far less. Most platforms report frequency as an average, so some people will see your ads more often than the figure suggests.
| Audience | Early phase | Later phase | Stop |
|---|---|---|---|
| Basket or checkout abandoners | Up to 2 a day for the first 2 to 3 days | About 1 a day or every other day | At purchase, or day 7 |
| Product viewers | About 1 a day for a few days | 2 to 4 a week | At purchase, or day 14 |
| Category browsers | 2 to 3 a week | 1 to 2 a week | Day 30 |
| Content readers and B2B visitors | 1 to 3 a week | About 1 a week | Day 90, or when the deal closes |
| Repeat-purchase buyers | None straight after purchase | 1 to 3 a week near the expected repurchase date | When they buy again |
This pattern is called tapering: front-load reminders while intent is fresh, then reduce them as interest fades, and stop entirely at purchase or when the window ends. Treat every number in the table as a starting point to test against your own click-through and conversion data, not a rule.
Explicit frequency caps vs indirect frequency management
An explicit cap is a setting that limits impressions per person. Indirect management means controlling frequency through budget, audience size and window length when no cap is available. Which you get depends on the platform, the campaign type and the objective:
- Google Ads: frequency capping is available in Display and Video campaigns, per day, week or month. Display caps count only viewable impressions, Video caps can limit impressions, views or both, and Google says frequency capping isn't supported in Demand Gen campaigns.
- Meta: its help centre documents two frequency controls for reservation campaigns: a frequency cap, such as two impressions every seven days, and target frequency, an average weekly frequency for campaigns of seven days or longer. For sales and conversion campaigns, plan on managing frequency indirectly.
Controls change regularly, so check the current options in each platform before you launch.
How budget and audience size drive frequency
Where there's no cap, budget and audience size set frequency for you. A small audience with a large budget will see the same people again and again, so match budget to audience size.
Impressions ≈ budget ÷ cost per 1,000 impressions (CPM) × 1,000
Average frequency ≈ impressions ÷ people reached
→ A $70 weekly budget at a $10 CPM buys about 7,000 impressions
→ Spread across 1,000 reachable people, that's about 7 a week each
→ Spread across 5,000 people, it's about 1.4 a week each
Hypothetical figures, assuming the full budget spends. If frequency is too high for a small audience, lower the budget, widen the window slightly, or combine audiences.
How to spot ad fatigue
- Click-through rate falls as frequency rises, week on week, for the same audience and creative.
- Cost per result climbs without a change in bids, targeting or seasonality. Meta's Ads Manager flags this as "Creative limited" or "Creative fatigue" when an ad's cost per result rises well above your past ads.
- Negative feedback grows, such as people hiding or reporting ads, or negative comments.
When those appear, reduce frequency, shorten the window or refresh the creative. The same logic of diminishing returns from repeated contact applies to email, as covered in holiday email volume and fatigue.
Who should you exclude from retargeting?
Exclusions do two jobs at once: they stop retargeting annoying people, and they stop it taking credit for sales that were already done. Most accounts need more of them than they have.
- Buyers of the advertised product. This is a product-specific exclusion: someone who just bought a sofa shouldn't see that sofa. Keep them excluded for as long as a repeat purchase is unlikely, which could be years for a mattress or a few weeks for coffee. If they bought a gift, they may never need it themselves.
- All existing customers, but only in campaigns whose job is acquiring new ones. In cross-sell or replenishment campaigns, customers are the audience, so use a separate campaign rather than a blanket exclusion.
- Converted leads, once they've submitted a form or booked a demo, so sales conversations aren't interrupted by "book a demo" ads. Syncing CRM status through regular customer list uploads keeps this current.
- Employees, who visit the site constantly. An uploaded employee email list, where platform terms allow, or an exclusion for internal-only pages handles most of them.
- Job seekers and support or login visitors who visited only careers, support, login or account pages. Exclude visits limited to those pages, not everyone who touched them, because logged-in customers who then browse products may still be relevant.
- Very short visits, such as people who left within a few seconds. Use this one with care: time-based rules depend on tracking and consent, quick mobile visits aren't always low intent, and a narrow rule can shrink audiences more than expected. Test it on one campaign before applying it everywhere.
- People without consent to advertising cookies where consent is required, and people who have opted out of marketing where that applies.
- Sensitive categories, such as health-related browsing, where platform rules and privacy law restrict or prohibit retargeting.
Purchaser exclusions only work if purchase data reaches the platform quickly and completely. If the conversion event is missing, delayed or blocked, the buyer stays in the "interested" audience and keeps seeing ads for something they own, which is one of the main causes of post-purchase retargeting. Phone, in-store and marketplace sales need separate uploads, and the basics are covered in how to set up conversion tracking correctly. Server-side tracking and consent-aware measurement, covered in consent mode v2, help keep that data flowing within the rules.
The principle Good retargeting is a reminder, not a shadow. The settings that make it less annoying, shorter windows, lower frequency and proper exclusions, are the same ones that stop it taking credit it hasn't earned.
Sequence the message instead of repeating it
Seeing the same ad ten times is irritating; seeing a short, logical sequence can be genuinely helpful. For a basket abandoner, a simple sequence might start with a plain reminder of what they left, move on to reassurance, such as reviews, delivery times or the returns policy, and finish with help or an alternative, such as a related product or a chat link. A discount, if you use one at all, belongs at the end of the sequence, not the start; leading with money off trains people to abandon baskets on purpose.
Here's what that might look like in practice for someone who left a pair of headphones in their basket. Day 1: "Still thinking it over? Your headphones are waiting," with the product image. Day 2 or 3: a short customer quote about battery life and comfort, plus "Free returns within 30 days." Day 4 or 5: "Questions about fit or compatibility? Chat with our team," linking to help. Day 6 or 7: a final, low-key reminder, or nothing at all. Then the person leaves the audience. Four different messages, each doing a distinct job, feel like service; the same "Don't miss out!" ad shown twelve times feels like being followed.
For product viewers, rotate between the product, social proof and a key benefit. For category browsers and content readers, use helpful content and best-seller collections rather than direct product pushes. Refresh creative regularly, and test which messages actually move people using the approach in the fundamentals of ad creative testing. And coordinate with your other channels: someone receiving abandoned-basket emails and texts doesn't also need a heavy run of ads saying the same thing, a coordination problem covered in orchestrating cross-channel flows without overlap.
Prove it's adding sales
The only reliable way to know what retargeting adds is to compare people who could see the ads with similar people who couldn't. That's a holdout, or lift, test: randomly withhold ads from part of the audience for a few weeks and compare conversion rates. Several ad platforms offer built-in lift studies, and geographic or time-based tests can work too.
A retargeting campaign spends $5,000 and the platform reports 500 conversions at an average order of $80: $40,000 in attributed revenue, a return on ad spend of 8×.
A holdout test shows the exposed group converts at 5%, while the matched group that saw no ads still converts at 4%.
→ Only (5% − 4%) ÷ 5% = 20% of those conversions were caused by the ads
→ Incremental conversions: about 100, worth about $8,000
→ Incremental return on ad spend: about 1.6×, not 8×
That doesn't mean retargeting failed. At 1.6×, it may still be profitable. But it changes how much you should spend, and it shows that most of the reported value belonged to the customer's own intent.
Results vary widely by brand, audience and settings; that's exactly why it's worth testing your own.
Lift tests can show that tighter windows, lower frequency and better exclusions reduce reported conversions while barely changing incremental ones, which means the same real result for less money and less annoyance. That's the clearest argument for running retargeting this way. Keep budgets stable during tests, following setting and scaling ad budgets, and give them long enough to reach meaningful numbers.
Respect consent and privacy
Retargeting depends on tracking people across visits, which makes it one of the most privacy-sensitive things marketers do. In many places, including the UK and EU, advertising cookies and similar technologies generally need consent, and India's DPDP Act is moving data use towards consent too. Retargeting audiences are also shrinking as browsers, operating systems and consent choices limit tracking, a shift covered in life after third-party cookies.
The practical response is to respect choices and rely more on data people have chosen to share. Honour consent properly, as covered in setting up cookie consent without killing your analytics. Build owned audiences, such as email subscribers and customers who've logged in, and use them carefully for re-engagement, following building a first-party data strategy. And avoid retargeting based on sensitive browsing altogether. Smaller, consented audiences, handled thoughtfully, often perform as well as larger ones handled carelessly.
A note on B2B retargeting
Retargeting works differently when the buyer is a committee rather than a person. Decisions take months, several people visit the site, and the person who reads the pricing page may not be the one who signs. Longer, lower-frequency windows suit that rhythm, and the message should shift from "come back and buy" to helping each person make the case internally: case studies, security and compliance information, implementation timelines and answers to common objections. Combined with account-level targeting, as covered in building an account-based marketing program, retargeting becomes a way to keep a whole buying group informed rather than chasing one visitor.
Retargeting at peak
The holiday season concentrates every retargeting problem. Many shoppers are buying gifts, so retargeting someone with the item they just bought for a relative is especially pointless. Budgets are higher, so frequency can climb fast without anyone noticing. And everyone is retargeting the same shoppers at once. Shorten windows for basket abandoners, tighten purchase exclusions, refresh creative more often, check frequency weekly and shift messaging as delivery deadlines approach, from offers to delivery dates to gift cards. After Christmas, past gift buyers may be a good audience for their own purchases, not a repeat of what they gave.
A quick audit checklist
- Are windows set by intent and buying cycle, not platform defaults?
- Is frequency capped or controlled through budget and audience size, and checked weekly against click-through and cost per result?
- Are buyers of the advertised product excluded for a realistic repurchase period, and all customers excluded from acquisition campaigns?
- Are converted leads, employees, job seekers and support visitors excluded?
- Is conversion data reaching the platforms quickly enough for purchase exclusions to work?
- Is creative sequenced and refreshed, with discounts used sparingly and late?
- Are retargeting ads coordinated with email and text follow-ups?
- Is consent respected, and is sensitive browsing excluded?
- Has a holdout or lift test measured incremental results in the last six months?
- Is someone responsible for reviewing frequency, negative feedback and hidden-ad rates every week during busy periods?
- Were windows, caps and exclusions adjusted after the last test, rather than left at their launch settings?
Before changing anything, record a baseline of current spend, reported conversions, frequency and click-through rate, so you can see what each change actually does.
Brands that want their retargeting rebuilt around incremental results rather than reported ones often work with a specialist performance marketing team.
The short version
Retargeting is effective and widely resented, and its reported numbers usually overstate what it adds because it targets people who were already likely to buy. Good retargeting is a reminder, not a shadow. Set windows by intent and buying cycle: roughly one to seven days for basket abandoners, three to 14 days for product viewers, two to four weeks for category browsers, and up to 90 days, gently, for content readers and considered B2B purchases. Remember that a window only makes people eligible; budget and frequency decide what they actually see. Keep frequency lower than instinct suggests, front-loaded after high-intent actions and then tapered, and watch for falling click-through, rising costs and negative feedback. Exclude buyers of the advertised product for their repurchase cycle, all customers from acquisition campaigns, converted leads, employees, job seekers, support visitors, people without consent and sensitive categories, and keep conversion data flowing so exclusions work. Sequence messages from reminder to reassurance to help, with discounts used sparingly and late. Run holdout tests: in our illustration, a reported 8× return became 1.6× incremental.
Is your retargeting helping, or just following people?
We rebuild retargeting around sensible limits and proven incremental results.
Explore Performance Marketing →Frequently asked questions
How long should a retargeting window be?
As long as the person's interest is likely to last, which depends on what they did and what you sell. As starting points, use roughly one to seven days for basket abandoners, three to 14 days for product viewers, two to four weeks for category browsers and 30 to 90 days for content readers and B2B pricing or demo visitors. Higher-priced and considered purchases justify longer windows; impulse buys and time-limited offers need shorter ones. A window only makes someone eligible to see ads, so longer is not automatically better. Start with windows matched to your typical time to purchase, then test.
What is a good frequency cap for retargeting ads?
There is no single correct number, and any figures you see are starting points to test rather than benchmarks. Allow more contact in the first few days after a high-intent action, such as up to two impressions a day for a basket abandoner, then taper, and keep longer windows to a handful of impressions a week. Explicit caps depend on the platform: Google Ads offers them in Display and Video campaigns, and Meta documents them for reservation campaigns. Elsewhere, manage frequency through budget, audience size and window length, and watch for falling click-through, rising cost per result and negative feedback.
Who should be excluded from retargeting?
Anyone who has already done what you wanted, anyone who was never a real prospect, and anyone you should not be following. That means buyers of the advertised product for a realistic repurchase period, all existing customers in campaigns aimed at new customers, converted leads, employees, and people who visited only careers, support, login or account pages. It also means anyone who has not consented to advertising cookies where consent is required, people who have opted out of marketing, and people associated with sensitive categories such as health conditions. Purchaser exclusions only work if conversion data reaches the platform quickly.
How do you know if retargeting is actually working?
By measuring incremental results, not just attributed ones. Retargeting targets people who have already shown interest, many of whom would have bought anyway, so platform reports often credit it with sales it did not cause, which makes return on ad spend look unusually high. The most reliable way to test this is a holdout or lift test: randomly withhold ads from part of the audience, then compare conversion rates between people who could see ads and people who could not. The difference is the incremental effect. Several ad platforms offer built-in lift studies, and you can also run geographic or time-based tests. Combine those results with sensible frequency, window and exclusion settings, and judge retargeting on incremental revenue and profit rather than platform-reported ROAS alone.
