In year one, B2B ads brought back 56 cents on the dollar: 0.56x.
The lowest honest number, not the last word.
Every dollar of 2025 ad spend brought back 56 cents, 0.56x in closed-won revenue we can trace back to an ad, divided by ad spend, inside 12 months. A customer cost $58,887, and 21.2% of the deals that reached a decision were won. Any plan that assumes advertising pays for itself inside the year is planning on a number this data does not produce. The scope: 153 advertisers, $57.6M of 2025 ad spend, 127 of them with revenue we could match back to their CRM.
Put closed-won revenue next to cost per lead on this week’s campaign sheet. Across 127 advertisers, every ad dollar brought back 0.56x, 56 cents of closed-won revenue we can trace back to an ad, inside 12 months, and one cut of the data beat break-even: LinkedIn campaigns aimed at 51–200-employee companies, 1.16x at $35,288 per customer, across 27 advertisers. Behind those numbers: $29.4M of lead-generation spend and 154,000 leads.
Before you quote 0.56x: three things it leaves out
- It counts only revenue we can trace back to an ad, so the true figure is higher. Revenue we could not tie to a deal an ad started counts as zero here, however real it was.
- Deals still open do not count. A deal that has not closed or died yet is not counted as a win, and advertisers with 6- to 18-month sales cycles had not finished closing their 2025 deals, so time alone moves this figure up.
- The spend side includes money whose results we cannot trace at all. Including budget spent on campaign types that never produce anything the CRM can see. Every one of those dollars divides the revenue we can trace.
2025 · 153 advertisers · $57.6M analyzed · 127 advertisers with revenue data · traceable payback · only groups with at least $50K spent · Methodology
Executive summary
- The money: across 127 advertisers, every dollar of 2025 ad spend brought back 0.56x, 56 cents. In closed-won revenue we can trace to an ad. A customer cost $58,887, and 21.2% of decided deals were won.
- The reporting call: label the number on every board slide. Counting every deal that any account we advertised to produced, the same spend reads 7.6x; counting only revenue we can trace back to an ad, it reads 0.56x. Both are real, and they answer different questions.
- The caveat: this is credit we can trace, not proof the ads caused the revenue, and it is deliberately the low end, three times over: traceable revenue only, no credit for deals still open, and divided by every dollar spent, including spend we cannot trace at all. Brand-building, word-of-mouth and long deals still running sit outside it, which is why we publish 0.56x as the lowest honest number rather than the whole story.
Playbook
- The play: put closed-won revenue and cost per customer on the same weekly sheet as cost per lead. Across this data every ad dollar brought back 0.56x, 56 cents of closed-won revenue we can trace back to an ad, inside 12 months, and no cost-per-lead report can see that.
- The setup: the one cut that paid for itself is narrow and copyable, LinkedIn aimed at 51–200-employee companies: 1.16x back, $35,288 per customer, 19.5% of decided deals won, across 27 advertisers.
- The measurement: count a customer only when the CRM says the deal closed, split the credit evenly across the ads that touched it, and never add up the conversions the ad platforms report.
Did B2B advertising pay for itself in 2025?
Not inside the year. Across 127 advertisers, every dollar of 2025 ad spend brought back 0.56x, 56 cents. Of closed-won revenue we can trace to an ad: $58,887 per customer, with 21.2% of decided deals won. One cut of the data beat break-even.
B2B advertising does not have a lead-price problem. It has a measurement problem. The cost per lead a team is judged on is an average of very different campaigns, and the return printed beside it is the ad platform’s own arithmetic rather than the CRM’s. Counted the careful way, credit split evenly across the ads that touched a deal, closed-won revenue only, and no one advertiser allowed to hold more than 40% of any result. The typical B2B ad program here is not a 3–5x engine. It brings back less than it costs in the first year, and most teams cannot see it, because nothing in platform reporting is built to show it.
Why is this lower than the return in your dashboard?
Because three different numbers get called ROI. The ad platforms count the conversions they claim. Influenced reporting counts every deal that any account you advertised to produced. This page counts only closed-won revenue on deals that an ad-generated lead started. Same 2025 spend: the influenced view reads 7.6x, the traceable view reads 0.56x.
Outside studies that ran proper tests found platform-reported returns roughly 2.3 times higher than the lift those tests could prove, and influenced pipeline two to four times hot. Those are other people’s studies, not our measurement, we cite them to place our number, not to produce it. The practical read: if your dashboard says 3x, most of the gap between that and this page is the definition, not the performance. So label the number every time. Ours is traceable payback, closed-won revenue we can trace back to an ad, divided by ad spend, inside 12 months, and we print the influenced view beside it rather than instead of it.
What does 0.56x not measure?
Three things, and all three would push the number up. It leaves out brand-building and word-of-mouth demand that never shows up as a lead record; deals still open at the 12-month cut-off; and any revenue an account you advertised to produced without a deal we can trace to an ad. Credit we can trace is not proof the ads caused it.
So this is a starting point, not a target. It measures the revenue we can trace inside the year against the money spent in 2025, credit split evenly, with one advertiser removed for holding too much of the result, and every group held to the same rules. It does not measure total business impact, profit, lifetime value, or what the ads caused, and a team whose sales cycle runs 6 to 18 months should expect its own traceable figure to lag its real economics. 0.56x is the number a CFO can pick apart downward; the influenced 7.6x is the one a marketer can argue upward.
How could this number be wrong?
Five ways, and we print all five. A rival analyst would say: splitting credit evenly undercounts wins only one ad touched; advertisers with 6- to 18-month sales cycles had not finished closing their 2025 deals; our 12-month window is shorter than the sales cycles we cite; only 127 advertisers have revenue we could match; and we removed the advertiser that held most of the wins.
Three of those five are right, and the direction matters: four of the five push the published number down, not up. Every cautious choice is disclosed and applied the same way to all 127 advertisers, which is what makes the number usable rather than flattering, credit split evenly, a close rate counted only on deals that actually closed or actually died, a 40% limit set in advance on how much of any result one advertiser may hold, and the removal that cost us our headline number, which is now in the kill list instead of the report.
The harder charge is who is in the data: advertisers who use Metadata and who keep their CRM tidy enough to match. True, and it cuts against us here. Weighted by how much each advertiser spent, a LinkedIn lead in this data cost $202. Cheaper than most of our own 88 LinkedIn advertisers actually paid. We are not showing the weak end of the market; we are showing that the figure the industry quotes as a target is beaten by a minority of advertisers spending a minority of the money. And on cause we concede the point outright: this is credit we can trace, not proof the ads caused the revenue. Then the burden flips twice. If your payback beats 1x, say which of our four cautious rules you beat. And show us a dataset joined to a CRM, credit split evenly, every deal counted once, that says otherwise.
Which setup actually paid for itself?
One, narrow and copyable: LinkedIn campaigns aimed at 51–200-employee companies brought back 1.16x at $35,288 per customer, with 19.5% of decided deals won, across 27 advertisers. Every other group we published came in under break-even, including the 0.56x across all 127 advertisers.
Cost explains most of it. Winning a mid-market customer in the 501–1,000-employee band cost $130,468 across 19 advertisers, against $35,288 in the small-company group, roughly 3.7 times as much for the same work, while the price of a lead barely moved between the ends of the size range ($241 at 51–200 against $204 at 1,001–5,000). That is the practical shape of it: fund the band that pays for itself as acquisition, with payback goals, and fund the bands that do not as influence, with coverage goals. Same budget, two scoreboards.
What should you change?
Rebuild the plan on traceable payback before the next budget cycle: 0.56x and $58,887 per customer across 127 advertisers, with the influenced 7.6x printed beside it, never instead of it. One labelled number per slide, and the conversions the ad platforms report never enter the payback line.
Judge the expensive size bands on coverage for one quarter, not on payback. A customer cost $130,468 in the 501–1,000-employee band and $80,109 at 1,001–5,000. Asking those bands to pay for themselves is the wrong scoreboard, so judge them on how many of the right accounts they reach.
“Advertising does not work.” 0.56x is a starting point, not a target, and it is the low end on purpose, since brand-building and deals still running sit outside it. Open the strategic lens and read the individual numbers, not the average.
Wire closed-won revenue back into the campaign sheet this week: deals started, won, lost and revenue per campaign, with the credit split evenly across the ads that touched each deal. If a campaign cannot report a won deal, it cannot report a return, and cost per lead alone would have called this data healthy at $202 a lead on LinkedIn.
Split one LinkedIn campaign down to 51–200-employee companies and watch it on its own for a quarter. That is the one cut that beat break-even, 1.16x at $35,288 per customer and a 19.5% close rate, across 27 advertisers, and running everything on one blended budget is how it stays invisible.
“Cheaper leads will fix the return.” In the same data the audience with the higher cost per lead produced more customers per 1,000 leads at a lower cost per customer, see the audience finding. Bidding the lead price down moves the number you report, not the revenue you book.
What are the exact numbers?
Six numbers on every cut of the data where we have them. Nothing is ranked on cost per lead alone: a cost number never appears without a result number beside it. A dash means the number is in the explorer rather than on this page.
| Published cut | Advertisers | Cost per lead | Customers per 1,000 leads | Close rate | Cost per customer | Traceable payback | Average won deal |
|---|---|---|---|---|---|---|---|
| All advertisers with revenue data | 127 | — | — | 21.2% | $58,887 | 0.56x | — |
| 51–200 employees × LinkedIn | 27 | — | — | 19.5% | $35,288 | 1.16x | — |
| Metadata-built audiences | 104 | $217 | 3.56 | 19.2% | $60,896 | 0.43x | $26,064 |
| Native platform targeting | 72 | $181 | 2.60 | 17.4% | $69,705 | 0.71x | $49,778 |
| LinkedIn (all sizes) | 114 | $202 | — | — | $63,312 | 0.57x | — |
| Image ads | 121 | — | — | 24.3% | $59,488 | 0.50x | — |
| 501–1,000 employees | 19 | — | — | — | $130,468 | — | — |
| 1,001–5,000 employees | 23 | $204 | — | — | $80,109 | — | — |
Traceable payback = closed-won revenue we can trace back to an ad, credit split evenly, divided by ad spend, over the 12 months after the 2025 spend. Close rate = we counted only deals that actually closed or actually died. Cost per customer = spend divided by the customers credited to it. Ad numbers come from the full $57.6M; revenue numbers from the $29.4M of lead-generation spend and its 154,000 leads matched to a CRM. Numbers not quoted here are in the interactive explorer, including the ones we could not publish, each labelled with the rule it failed.
How do we know?
Every number above passed the same publication gates before it reached this page. In brief:
- How many advertisers it takes to publish: at least 5 advertisers behind any published ad number and at least 8 behind any revenue number, plus at least 3 closed-won deals. The all-advertiser revenue row rests on 127 advertisers; the 51–200 × LinkedIn row on 27.
- No one advertiser may swing a number: no single advertiser above 40% of a published revenue number’s spend or above 40% of its closed-won deals (50% for ad-only numbers). One advertiser failed that test across the whole dataset and was removed, its deals and its matching spend. Which is what moved this number.
- Spend floor: at least $50,000 spent behind any published number.
- Scope and window: $57.6M of 2025 ad spend across 153 advertisers; revenue counted on the $29.4M of lead-generation spend and its 154,000 leads matched to a CRM, over the 12 months after the spend. The other windows we considered, 6 to 18 months matched to the sales cycle, 180 to 365 days, 90 days with a warning that it undercounts, are printed in the methodology.
- Credit rule: we split the credit evenly across the ads that touched the deal, count closed-won revenue only, and count every deal once. The revenue credited to marketing can never exceed the revenue in the CRM. The conversions the ad platforms report are a check here, never a result.
- What kind of evidence this is: every figure describes what happened in 2025, not a controlled test. No holdout groups, no randomization, nothing here proves the ads caused the revenue.
- What else could explain this: deals that had not finished closing, and who is in the data. Advertisers whose buying cycles run 6 to 18 months still had 2025 deals open at the cut-off, so this figure can only rise as those land; and the data is advertisers whose CRM records match cleanly, which favours teams that can measure this at all. Both are named, neither is controlled. Which is why we publish this as the lowest honest number.
Full rules in the methodology, including what we refused to publish, with the exact reason each draft claim was killed. New here? Start with typical vs best-case: how to read any benchmark.
Cite as: Metadata 2026 B2B Ad Spend Benchmark (153 advertisers, $57.6M, 2025), metadata.io/benchmark-report-2026
What do you do different tomorrow morning?
One named change per seat. The columns reorder to match your reading lens; both are always on the page.
Rebuild the board number
The one change: replace “marketing ROI” on the board slide with two labelled numbers, traceable payback and influenced pipeline.
- Ask for your own trailing payback counted the careful way: credit split evenly across the ads that touched the deal, closed-won only, deals still open reported separately. Put it next to this data’s 0.56x and $58,887.
- If what comes back is a number the ad platforms reported, that is finding number one. Closing the gap between the two is the work, and outside studies put platform reporting roughly 2.3 times hot.
- Split the scoreboard by company size: bands that pay for themselves get acquisition goals, bands at $130,468 per customer get coverage goals. Same budget, two scoreboards.
Make closed-won the contract
The one change: add deals started, won, lost and revenue columns to the weekly campaign sheet, with the credit split evenly across the ads that touched each deal, before you touch a bid.
- Recount your close rate on decided deals only, deals that actually closed or actually died. Counting deals still open is what makes weak campaigns look like closers; this data reads 21.2% under the honest count.
- Rank campaigns on customers per 1,000 leads and cost per customer, not cost per lead. The audiences we built here cost $217 per lead against $181 for the platforms’ own targeting, and still produced more customers per 1,000 leads.
- Copy the one cut that paid for itself before you optimize anything else: LinkedIn aimed at 51–200-employee companies, 1.16x at $35,288 per customer, across 27 advertisers.