Open your ad account and look at your biggest retargeting audience. Now ask one question: when did most of these people actually visit your site?
For a lot of accounts, the honest answer is “months ago.” The audience was set up once, given the longest window the platform allowed, and never touched again.
It still looks healthy in the dashboard because the count keeps growing. But a large share of the people inside it stopped caring about you a long time ago. Some bought from a competitor, some were never buyers at all, and a few were your own employees.
That is what a stale retargeting audience looks like, and it quietly drags down every campaign that uses it. Here is why it happens, how to check whether it is happening to you, and how to rebuild the audience around behavior that is actually recent.
How retargeting audiences go stale
The three big platforms each hand you a lookback window when you create a website audience. Meta lets you keep a website visitor in a Custom Audience for anywhere from 1 to 180 days. Google Ads defaults data segments to 30 days but lets you stretch membership to 540 days on both Search and Display. LinkedIn gives you four options for website retargeting: 30, 60, 90, or 180 days.
Those maximums exist for edge cases, not as recommendations. But in practice, the longest window gets picked because it makes the audience bigger, and bigger feels safer. The result is a pool where someone who read one blog post in March is treated identically to someone who hit your pricing page yesterday. On Google’s 540-day maximum, that March visitor could be from the previous year.
There is a second, less obvious way audiences rot: the rule that defines them breaks. Most website audiences are built on “URL contains” logic. Rename a product page, migrate to a new domain, change your URL structure during a redesign, and the rule stops matching. The audience does not throw an error.
It just stops adding new people while the old ones age out or, on a long window, sit there.
And a third: nothing removes people who no longer belong. Customers who already bought. Leads your sales team disqualified. Job seekers, vendors, and competitors doing research.
Unless you built exclusions, they are all still in the pool, and you are paying to show them ads.
Recency is not a nice-to-have. It is most of the value.
This is not a hunch. In a large field experiment published in the Journal of Marketing Research, researchers from Stanford GSB and Santa Clara found that retargeting caused 14.6% more users to return to the advertiser’s site within four weeks, but the effect shrank the longer it had been since the person’s visit. Ads in the first week after a visit did more work than ads shown in later weeks. The intent you are chasing is front-loaded.
We covered the same pattern from a different angle in our post on intent decay: the signal a visit gives you is strongest in the first two days and fades fast after that. A 180-day window is not “more reach.” It is a small amount of fresh reach plus a large amount of expensive noise.
One caveat, because B2B people will raise it. Long sales cycles are real. 6sense’s 2025 Buyer Experience Report puts the average B2B buying cycle at roughly 10 months, down from about 11 the year before. So yes, someone who visited six months ago might still be in-market.
But that does not mean they should get the same ad, at the same frequency, at the same bid as someone who visited this week. It means they belong in a different tier. More on that in a minute.
A 15-minute audit to find out if yours is stale
Do not guess. Check.
- Read the window on every active audience. Write down the lookback setting for each. If most of them sit at the platform maximum, you already have your answer.
- Check the rules against the live site. Open each URL-based rule and confirm the URL pattern still exists. Redesigns and migrations break these silently. If your site changed in the last year, assume at least one rule is dead until proven otherwise.
- Split by recency and compare. On Meta, duplicate your main audience into 0–7, 8–30, 31–90, and 91–180 day versions (exclude the shorter from the longer so they do not overlap). On Google, create segments with different membership durations. Run them side by side for two weeks and compare CTR, conversion rate, and cost per conversion. In most accounts, the gap between the freshest tier and the oldest is not subtle.
- Look for people who should not be there. Are converters excluded? Closed-lost leads? Current customers? Your own office IP range? If the answer is “I think so,” go verify.
- Check the platform-level changes you may have missed. Google retired the Privacy Sandbox APIs that were supposed to replace cookie-based remarketing, with removal targeted for Chrome 150 in July 2026, while third-party cookies themselves stayed. If you built anything on Protected Audience, it is gone. Our plain-English cookie update has the full timeline.
How to fix it: rebuild around recent behavior, not pixel age
Tier by recency and treat the tiers differently. The 0–7 day group gets your most direct offer, the highest frequency cap you are comfortable with, and the most aggressive bid. The 8–30 group gets proof: case studies, comparison content, a reason to come back.
Anything older than 90 days should either be cut or moved into a low-frequency “stay in view” campaign with a small budget. Do not let it share a budget with your fresh traffic, because the algorithm will happily spend on the cheap, unresponsive tail.
Define audiences by what people did, not just that they showed up. A pricing page visit, a demo page visit, a comparison page visit, or a return trip within a week all say something. A single blog read does not.
Build separate audiences for high-intent pages with short windows (7–30 days), and put low-intent pages on a longer window or leave them out of retargeting entirely. We walked through this logic in our behavioral segmentation post and it applies directly here.
Build real exclusion lists and keep them synced. Converters, customers, and disqualified leads all come out. Upload customer lists from your CRM on a schedule, not once. A retargeting audience with no exclusions is not a retargeting audience; it is a list of everyone.
Stop relying on the browser to remember people. Pixel-only audiences depend on a cookie surviving on the visitor’s device. Cookies get cleared, blocked, and expire; first-party data does not.
If you know who a visitor is, from a form fill, a login, a CRM record, or an identity resolution layer, you can build the audience on that identity and refresh it from your own systems. For B2B sites where most traffic never fills out a form, this is the piece most teams are missing.
It is what Smart Pixel is built for: identifying who is actually on the site so the audience is defined by the person and their recent behavior, not by how long a cookie has lasted. Use it if that is your gap. If your site already captures identity, you do not need it for this.
Put a review date on the calendar. Quarterly is enough. Re-check the rules, the windows, the exclusions, and the recency split.
Audiences are not set-and-forget infrastructure. They are a living definition of who you think is in-market right now, and “right now” moves.
The short version
If your retargeting audience is a single pool on a maximum lookback window with no exclusions and rules you have not opened since the site was redesigned, it is stale. You are paying to reach people who have moved on and diluting the results from the people who have not.
Split by recency. Define by behavior. Exclude who does not belong. Anchor to identity where you can, then check it again in 90 days.
Not sure where your audiences actually stand? Our Traffic Intelligence Evaluation looks at who is really visiting your site and how much of that traffic your current setup is missing. 14 days. Your actual visitors. No obligation.
