Metadata 2026 B2B Paid Media Benchmark Report: an annotated summary

The short version

The report's headline claim is that creating demand from cold audiences costs about the same per lead as retargeting people who already know you: $187 against $196 [1][3]. Its second claim is that the typical B2B advertiser puts 25% of budget into creating demand, 7% into retargeting and 16% into buying clicks [1][4]. Its most alarming number is that $12.7M of the analysed spend went to click or traffic-objective campaigns and 99.4% of those campaigns recorded no lead at all [1][5]. Everything else in the report is evidence underneath those three points.

The dataset is 153 B2B advertisers, $57.6M of calendar-2025 spend and 211,000 leads across LinkedIn, Google Ads, Facebook and Instagram; Reddit and Microsoft Ads spend is in the totals but too thin to publish as segments [1][2]. Pipeline metrics use a $29.4M subset of lead-generation campaigns with CRM opportunity attribution, covering 127 advertisers and 154,000 CRM-joined leads [1][2].

The ten findings, with links to each insight page

What each finding says, and what it does not

Finding 01: create vs capture, $187 vs $196 per lead

The report's central argument is that creating demand is not the expensive half of a paid program. Cold-audience campaigns produced a lead for $187 while retargeting produced one for $196, which the report calls "close enough to call a tie, not the premium most plans budget for" [1][3]. The bigger gap it finds is by ad format inside cold audiences: a document ad produced a lead for $148 against $358 for video [1][3]. It also notes that on the same cold audience, Meta bought the cheaper lead and LinkedIn converted more of its clicks into leads [1].

Our annotation: this is a cost-per-lead comparison, and cost per lead is the metric the report's own finding 09 warns against using alone. Read 01 and 09 together. The credible version of the claim is that cold and retargeting cost about the same per lead, and the report has 153 advertisers behind that. Whether cold leads close at the same rate as retargeted leads is answered only partially, in the audience yield page. Our own create vs capture demand page collects other sources on the same question.

Finding 02: budget allocation, 25% / 7% / 16%

Across 109 advertisers, the typical split was 25% of budget creating demand, 7% retargeting and 16% buying clicks, with the remainder in objectives the report does not classify into those three lines [1][4]. The insight page publishes the bottom quarter, the middle and the top quarter for each line rather than a single average, and states plainly that "it is not a recommended mix" [1][4].

Our annotation: this is the most honest allocation benchmark we have indexed, precisely because it refuses to recommend. Most vendor allocation charts are a pitch for the vendor's product category. The weakness is that the three lines do not sum to 100%, so a reader has to work out what the remaining budget is doing. Compare with our budget allocation benchmarks page for other published splits.

Finding 03: $12.7M on click campaigns, 99.4% with no lead

Traffic-objective campaigns took 16% of the typical budget and 22% of all analysed spend, across 112 advertisers, 61 of whom put more than 30% of budget there [1][5]. The lead report shows 66 leads against $12.7M, which computes to $191,686 per lead, but only four advertisers recorded a single one [1][5]. The report's framing is precise: "That is not a price. It is $12.7M with no lead capture wired to it, and nobody can say what it bought" [1].

Our annotation: this is the finding most likely to be misquoted. It does not show that click campaigns are wasted; it shows that lead reporting cannot see them. The report's recommendation, cap traffic campaigns, give each a stated brand purpose and remove them from cost-per-lead reporting, is sound. The 99.4% figure is a campaign-count share, not a spend share, which the insight page makes clear and secondary summaries often do not.

Findings 04 and 05: document ads and lead-gen forms on LinkedIn

LinkedIn document ads produced a lead for $142, cheaper than image ($200) or video ($265), converting 11.9% of clicks into leads [1][6]. The report says most advertisers skip the format, and that its pipeline numbers "were killed by the concentration rule," so it should be bought on lead economics rather than a promised return [1][6]. On the same offer, LinkedIn's in-feed lead-gen form cost $193 per lead against $346 for a landing page, a 44% gap from one campaign setting [1][7].

Our annotation: these two are the findings a demand-gen manager can act on this week, and the forms finding matches the direction of other public LinkedIn benchmarks we track on LinkedIn Ads benchmarks. The caution the report attaches to document ads (no pipeline number) is exactly the kind of caveat a vendor report usually omits.

Finding 06: the $202 CPL illusion

The $202 cost-per-lead figure widely quoted as a B2B benchmark is, in this dataset, a number three out of four advertisers failed to reach; most paid more [1][8]. The report argues that teams are being measured against a bar the market never cleared, and points readers at the Reality Gap calculator to find their own position [1][8][13].

Our annotation: this is a fair point about blended averages, and it is the reason our CPL by industry page reports ranges. The $202 figure's provenance is not fully spelled out on the insight page; treat "everyone quotes" as rhetorical.

Findings 07, 08, 09: Instagram, the CAC cliff and audience yield

Instagram produced the cheapest lead in the dataset at $138, on a small slice of budget, and the report cannot yet say what those leads became [1][9]. Cost per customer on LinkedIn ranged from $35,288 targeting 51 to 200-employee companies to $130,468 at 501 to 1,000 employees, easing to $80,109 at 1,001 to 5,000; the small-company band returned 1.16x [1][10]. The pricier of two audiences produced 3.56 customers per 1,000 leads against 2.6, with CAC of $60,896 against $69,705 and return of 0.71x against 0.43x [1][11].

Our annotation: 08 and 09 are where the report earns its keep. A blended cost-per-customer figure hides a near four-fold spread by company size, and the audience yield page shows a CPL dashboard actively defunding the better audience. These pipeline figures come from the stricter $29.4M CRM-matched subset with at least 8 advertisers and 3 closed-won opportunities per published group [2], which is a tighter bar than any other vendor benchmark in our index applies. See B2B CAC benchmarks for how these compare to other public CAC figures.

Finding 10: the 0.56x payback floor

Ads returned 0.56x in year one, 56 cents on the dollar, at $58,887 per customer and a 21.2% close rate on the $29.4M matched subset [1][12]. The report is explicit that this is "revenue we can trace back to an ad, not proof the ads caused it, a starting point, not a target" [1][12]. It also describes its attribution rule: split credit evenly across ads that touched the deal, count only deals that closed or died, state one window, and never let a platform's own conversion count into the return [1][12].

Our annotation: a vendor publishing a sub-1x first-year return on its own platform's spend is unusual, and the reason is on the report page itself, in a comparison of a "122X" customer case study with the 0.56x floor: "Different measure, different group, different job" [1]. That paragraph is the best short explanation of benchmark versus case study we have seen in a vendor report.

Methodology and the kill list

Media metrics (CPL, CPC, CVR) publish only from segments with at least 5 advertisers, $50,000 of combined spend and no single advertiser above 50% of segment spend [1][2]. Pipeline metrics (CAC, traceable payback) require at least 8 advertisers, at least 3 closed-won opportunities and no single advertiser above 40% of the group's wins, a rule the report says was written before the numbers were seen [1][2]. Groups that fail are withheld with a stated reason rather than estimated or blended into a parent [1][2]. The report keeps a public kill list of eight claims it chose not to publish, and a limits section on what the data cannot answer, including whether the revenue would have arrived anyway [2].

The data is downloadable as CSV and JSON under CC BY 4.0, with suppressed cuts absent rather than zeroed, and as a full PDF plus CMO and demand-gen editions [1][14][15][16]. A reference table by channel lists CTR, CPC, CPM and CPL with the advertiser count on each row [1].

Our read of the limits: the sample is Metadata.io's customer base, which skews toward B2B SaaS running paid social and search through one platform. It is not a survey of the market. Pipeline figures depend on CRM attribution quality at 127 advertisers. And the report cannot separate correlation from cause, which it states. Those are the same limits we attach to every platform-derived benchmark in our methodology; this one is simply more explicit about them.

Frequently asked questions