LinkedIn publishes four numbers for one ad format. None of them agree

The correction that never needed making#

Start with the fox at the edge of a birch wood, sieve in one paw, shaking a press release until only the provable pieces stay in the mesh. That is the posture this week deserves. On 30 July 2026 PPC Land reported that LinkedIn had restated its Event Ads viewership benchmark, moving the figure from 31% to 31x. Anyone buying LinkedIn Ads should want that checked, because a headline claim moving from a percentage to a multiple is either a typo being tidied or a benchmark being rebuilt in public.

So we went down the burrow. The LinkedIn Newsroom post dated 10 November 2025 carries the wording “up to 31x* more viewership”, and an archived copy of the same page from 18 November 2025 carries it identically. The restatement did not happen. The 31x was already there in November 2025, and no LinkedIn-owned page carrying 31% turned up anywhere in the check.

That is the correction, stated plainly and stated once. Trade desks work fast on a thin beat, this one chased a real oddity to the wrong den, and folkfox has misread a source before and will again. The caveat cuts our way too: failing to find a page is not the same as proving no page existed. Absence of evidence is a weak floor to stand on, and we are standing on it. The LinkedIn Ads question underneath the correction is the one worth the daylight.

The bigger quarry was underneath. Checking one figure meant reading every surface LinkedIn uses to sell the format, and those surfaces do not agree with each other. At least four claims, one LinkedIn Ads format, three different shapes of evidence behind them, and no route a buyer can take to reconcile them. That is not a trade outlet's problem to fix. It is a disclosure problem, and it sits squarely with the platform.

Marketing benchmarks are load-bearing. The benchmark sets the forecast, the forecast sets the budget, the budget sets the headcount, and somewhere at the end of that chain a media buyer defends a number nobody ever got to inspect. We keep the same audit running across social media ad costs, whichever platform publishes the figure and however flattering it looks. LinkedIn Ads are worth buying on their merits. They are not worth buying on a footnote nobody has read.

There is a wider warning here for every buyer of paid social, not only for the teams running LinkedIn Ads. Platforms are the only party holding the log files, so the platform is always the primary source for its own performance. That is a permanent structural problem rather than a LinkedIn problem, and it means the discipline has to live on the buyer's side of the table.

Four numbers for one format: the LinkedIn Ads evidence trail#

Set the claims side by side and the trail forks. The newsroom says one thing, the product page says three more, the marketing blog says something else again, and each rests on a different kind of evidence. None of them cite each other. None of them share a measurement window. Read as a set they are not a benchmark, they are a brochure, and anyone planning LinkedIn Ads has to pick one and hope.

The headline number lives on the LinkedIn Newsroom: “Marketers who use Event Ads see up to 31x* more viewership, and brands, like Impact.com, have doubled event attendance when using the ad format.” Follow the asterisk and it resolves to five words, “Based on 2025 internal watch data”. That is the entire methodology. No sample, no control, no definition of a view, and no statement of what the multiple is measured against.

Four of those rows come from the LinkedIn Event Ads product page, a live surface selling the format today. It states that “LinkedIn Events have the ability to drive 131% higher CTR” compared with regular video, that “this lift in performance continues with a 19% higher engagement rate” after the event, and that the format “can drive up to 40% lower cost per registration”. A fourth line reports that event attendees are “1.5x more likely to engage with future company marketing” than non-attendees.

All four of those LinkedIn Ads claims are the platform's own uncontrolled figures, unaudited and unverifiable from outside, and all four hang off one shared footnote. That footnote is the most honest sentence on the page.

Rendered in British dates, that window opened on 1 May 2023 and closed on 18 June 2024. It has been shut for over two years, and the figures it produced are still on sale as current LinkedIn Ads performance. Nothing about that is hidden. It is printed on the page in grey type, which is exactly where a careful buyer should be looking and exactly where almost nobody looks.

Then comes a fifth set entirely. A customer case study on the LinkedIn Marketing Blog reports livestream views “increased by an incredible 4,024% compared to the previous year, reaching 397,000 views”, total minutes viewed up 399% to 20,000 minutes, and unique views up 3,119%. One account. One year-on-year comparison. No sample size, no control group, no comparison set. A fine result for that brand, and no kind of benchmark for anyone else.

Four surfaces, five sets of figures, and not one of them can be checked against another. A buyer briefing LinkedIn Ads who wants to know whether Event Ads outperform ordinary video has, on the platform's own published evidence, no way to find out. That gap is the story, and it is the same gap we tracked through clicks turning into visibility on the search side.

The footnote that closed in June 2024#

Take the pilot-tester footnote seriously, because it does more work than any sentence on the page. A select group of pilot testers is a self-selected sample. Advertisers who volunteer for a new format are keener, better resourced and more attentive than the average account, so a pilot cohort compared with everyone else will flatter the format by construction. That is not fraud, it is selection, and every performance marketer has met it before.

The window matters as much as the group. Between 1 May 2023 and 18 June 2024 the auction those advertisers bought in was a different animal. Competitive density, creative norms, feed ranking and the price of attention have all moved since. Quoting a lift measured in that window as though it still holds in August 2026 asks a buyer to accept a two-year-old reading of a market that has repriced itself twice over. Anyone forecasting LinkedIn Ads this quarter is being handed a snapshot of a different auction.

The LinkedIn Ads cost claim is the one that moves money#

LinkedIn ad costs are the first question a finance director asks, and “up to 40% lower cost per registration” is exactly the shape of number that gets typed straight into a spreadsheet. “Up to” is doing heroic work in that sentence. It describes a ceiling reached by some accounts in a favourable pilot, not a median any buyer should plan against, and the difference between those two readings is the difference between a forecast that holds and one that quietly does not. Budget LinkedIn Ads against the ceiling and the shortfall arrives at week six.

None of this makes Event Ads a bad buy. It may well be an excellent one, and the fox has no quarrel with the format itself. The quarrel is that a buyer cannot tell, because the evidence offered is not the kind of evidence that answers the question being asked. Planning LinkedIn Ads runs on comparability, and four incomparable numbers give a planner nothing to compare.

Compare the standard we hold paid search to. A bidding change is expected to arrive with its mechanism, its affected campaign types and its date attached, and a planner handed “internal watch data” in place of that documentation would send it straight back. The tolerance is looser on the social side of the plan, LinkedIn Ads included, and it should not be.

What a properly footnoted number looks like#

Here is the sharpest scent in the whole thicket: LinkedIn already knows how to do this, and does it on the very same page. The LinkedIn Newsroom post that carries the 31x figure also states that “LinkedIn commissioned Censuswide to survey 3,251 B2B marketers (18+)” across thirteen markets, the UK, the USA, France, Germany, Spain, Brazil, the UAE, the Netherlands, Singapore, India, Australia, Italy and Sweden, with data “collected between April 22 - May 6, 2025”. Sample, panel, geography, fieldwork dates. Textbook.

The 31x figure is excluded from that methodology and footnoted separately to internal watch data. One page, two standards. Which means the thin disclosure around the LinkedIn Ads performance claims is a choice rather than an oversight, and choices can be changed by whoever owns the page.

Look at how other platforms handle an awkward number. Snap announced on 31 July 2026 that “wholly AI-generated videos will no longer be eligible for recommendation on Spotlight”, then said the quiet part in the same breath. Snap also reports unique Spotlight contributors up more than 120% year over year, which is its own uncontrolled figure with no baseline disclosed, so the discipline is not total. Naming your own limitation beside your own claim remains the move worth copying, and it is the same tension we picked apart in platform policy deciding what creative can do.

Roblox published an inconvenient number on 30 July 2026 and attributed it precisely. Bookings grew 8% and “landed at the low end of our guidance range”, which the company puts down to a decline in per-hour monetisation as players shift towards lower-monetising games, compounded by changes to its Recommended for You algorithm. It also reports that 57% of global daily active users have age-checked, again its own uncontrolled count with no outside verification attached. Precision of attribution is the point. A number with a named cause can be argued with.

Apple shows the opposite failure mode without a word of dishonesty. On 30 July 2026 it reported Services revenue of $30.7 billion, a June-quarter record, inside total revenue of $109.4 billion, up 16%. Apple does not break out App Store or advertising revenue, because Services is a blended line. The aggregate is disclosed, the composition is withheld, and every advertiser modelling that market is guessing at the mix.

Regulation is arriving for one kind of claim and not the other. The European Commission began enforcing AI Act Article 50 transparency obligations on 2 August 2026, with penalties reaching €15 million or 3% of global annual turnover, while advertising benchmarks stay entirely unregulated. Two asides make the same shape. The ASA held an advertiser responsible on 5 August 2026 for a landing page it did not control, because the test in law was who paid. And a Deepsee.io study of 4,346 AI companion apps rated 2,589 of them, 60%, accessible to minors, a finding published by a vendor that sells screening. Responsibility does not follow control, and interest does not follow evidence.

That last point deserves saying about ourselves too. Agencies publish numbers with an interest attached, and readers should discount ours accordingly. The remedy is not purity, it is disclosure, which is why every figure in the visibility reporting we build for clients carries its source and its date on the same row, LinkedIn Ads benchmarks included.

Five tests before a LinkedIn Ads benchmark reaches your plan#

None of this needs a crusade. It needs a sieve. Any figure a platform publishes can enter a plan, provided it walks past five questions first, and the questions take roughly four minutes per number. Run them on LinkedIn Ads, run them on Meta, run them on the retail media deck that arrived this morning. The fox does not care whose claim it is.

Then write the answers down. The discipline that matters is not scepticism, it is the paper trail. A plan that records where each assumption came from can be defended when the quarter misses, and a plan that cannot is confidence with a spreadsheet attached. Build the habit into the brand strategy document and the paid social strategy document alike, one line per benchmark, source and date sitting beside it. A LinkedIn Ads figure that survives that treatment has earned its place in the model.

The same sieve catches changes that arrive without being asked for, the sort we have picked over before in retail media placements. Platform defaults and platform benchmarks are the same species of decision: made elsewhere, delivered as fact, and yours to audit or inherit. A paid social strategy that audits both, LinkedIn Ads and everything sitting beside them, is a strategy that survives a difficult quarter.

Do that and four irreconcilable numbers stop being a trap. Each one enters the plan with its footnote showing or it does not enter at all, and the forecast carries a line saying which figures are load-bearing and which are decoration. LinkedIn Ads may still be exactly the right buy, and for a good chunk of B2B they probably are. The difference is that the case will be one you wrote, from evidence you inspected, rather than one the platform wrote for you.

If a benchmark in your current plan cannot survive those five questions, that is worth knowing before the board meeting rather than after it. Send us the deck and we will run the sieve over it: start the conversation and bring the LinkedIn Ads numbers you have been quoting. Some will hold. The ones that do not are better found in a quiet meeting than in a quarterly review, and the fox would rather forage in the archive than defend a figure in the open.

Frequently asked questions#