How do you tell if a LinkedIn ad is actually "winning" from the outside?

There's no public metric for an ad's return, but there is a reliable proxy: how long a company kept paying to run it. Nobody funds an underperforming creative for three months by accident, ad budgets get reallocated fast when something isn't converting, so observed duration is the closest thing to a public performance signal that exists in advertising. Across AdPeekly's dataset, the median campaign with a reported start date runs 19 days, and 11.8% run past 90. Those long-tail survivors are what "winning" means on this page: not a guess about creative quality, a measured fact about what companies kept investing in.

Does a longer-running ad always mean a better ad?

Not always, but it's the strongest signal available without access to a company's own performance dashboards, which no third party has. A campaign can run long because it's genuinely converting, because it's evergreen brand content nobody's actively managing, or occasionally because nobody got around to pausing it. What duration reliably rules out is the opposite case: a three-day test that never scaled is not a campaign anyone kept betting on. Treat long-running ads as a shortlist worth studying, not a guarantee, and combine duration with the company context (industry, funding stage) covered on the ad intelligence page to judge whether a given long-runner is actually relevant to your situation.

If a company runs many variants of one ad, does that mean it's a winner?

Not by itself, and it's a different signal from duration. Running many near-identical variants (a swapped headline, a cropped image, a new call to action) usually means active testing, which can happen early in a campaign's life just as easily as late in one. AdPeekly deduplicates variants down to one creative per distinct design specifically so variant volume doesn't get mistaken for popularity: the 266,919 published ads in this dataset collapse to 100,659 distinct creatives once that grouping is applied. The rule is at /methodology#deduplication. Duration on the deduplicated creative, not the raw variant count, is the stronger winning signal.

Use case

Steal Winning Ideas: Find the LinkedIn Ads Companies Keep Running

Ad libraries are built for transparency, not for finding what worked. LinkedIn's own tool shows you every ad currently live and tells you nothing about which ones a company kept funding for a reason. AdPeekly measures that instead: how long each of 266,919 published ads actually ran.

What counts as a "winning signal"

Two things, and both are measurable rather than guessed: how long a campaign ran, and how many creative variants a company kept producing around one idea. Neither one is a performance metric in the strict sense, nobody publishes conversion rates, but duration in particular is a strong proxy. Ad budgets get reallocated quickly when something stops converting, so a creative still running after three months is one somebody deliberately kept paying for, not one that slipped through unnoticed.

Across the 40,816 creatives in AdPeekly's dataset where LinkedIn reported a campaign start date, the median run is 19 days. The distribution is heavily front-loaded: a quarter of campaigns are gone within 8 days, but the top decile runs past 102. Those survivors, 7.4% of dated creatives running 61 to 90 days and 11.8% running past 90, are what this page means by "winning ads."

How to actually find them

Finding a long-running ad by hand means opening LinkedIn's Ad Library one advertiser at a time and guessing at how long each ad has run, since the tool doesn't show a start date at all. Sorting the AdPeekly dataset by observed duration turns that guesswork into a query: order by days running, and the survivors come to the top. That's the single most useful thing you can do with a dataset that remembers ads after LinkedIn's own tool has dropped them.

Format matters too, though less than intuition suggests. Carousel ads, one of the formats most associated with sustained testing and iteration, hold 2,174 distinct creatives in this dataset with a median observed duration of 23 days across 941 dated creatives, longer than several other formats but not dramatically so. The full breakdown, with real carousel examples and measured card specs, is at /linkedin-ads/carousel. For the complete duration breakdown by industry and funding stage, /stats/linkedin-campaign-duration is the source page every figure here is drawn from.

Stop guessing which ads worked. Sort by how long they ran.

Open the full app and sort by observed campaign duration, or connect the MCP server and ask an assistant you already use to pull the long-runners for you. Free tier covers 3 searches a day, no card required.

See pricing Get a free MCP key

No card for the free tier. Full tool reference at /mcp.

Thousands of LinkedIn campaigns