The audiences that cost 20% more per lead came with 37% more customers.
Cost per lead ranks audiences backwards.
The audiences we built produced 37% more customers per 1,000 leads at a 13% lower cost per customer. Native targeting brought back more revenue per dollar only because the deals it won were simply bigger, and we cannot tell whether that is the targeting or the advertisers who chose it, so we claim the cost-per-customer win, not the revenue one.
The audience that looked expensive was the cheap one. Across the lead-generation campaigns we can follow into a CRM, the subset every figure on this page is built from, the audiences we built cost $217 a lead against $181 for the ad platforms’ own targeting, and turned those leads into 3.56 customers per 1,000 against 2.60, at $60,896 a customer against $69,705. If your audience budget is set by lead price, it is being set by the wrong number. Across 104 advertisers on the audiences we built and 72 on the platforms’ own targeting.
Why this differs from the explorer. These figures cover the lead-generation campaigns with CRM attribution, because a cost per customer cannot be computed without one. The dataset explorer counts every campaign for the same audience, so it reports a slightly lower cost per lead, $208 against $217 for the audiences we build and $176 against $181 for platform targeting. Same advertisers, wider set of campaigns, not a correction.
Swap the cost-per-lead column in your audience sheet for two columns: customers per 1,000 leads and cost per customer. In 2025 the audiences we built won 19.2% of decided deals against 17.4% for the ad platforms’ own targeting, while costing $217 a lead against $181. Across 104 and 72 advertisers: the cheaper lead lost on every column that contains a customer.
2025 · 104 and 72 advertisers · $29.4M of lead-generation spend · 154,000 leads matched to a CRM · traceable payback · only groups with at least $50K spent · Methodology
Executive summary
- The money: the pricier audience made the cheaper customer. The audiences we built ran $217 a lead against $181 and produced 3.56 customers per 1,000 leads against 2.60, at $60,896 a customer against $69,705.
- The reporting call: a report ranked on cost per lead takes budget away from the audience that produces more customers, every quarter, without ever showing why. Rank on customers per 1,000 leads and cost per customer, and keep lead price as a check rather than a verdict.
- The concession: the platforms’ own targeting brought back more per dollar, 0.71x against 0.43x. Entirely because the deals it won were simply bigger: $49,778 against $26,064. We claim the cost-per-customer win, not the revenue one.
Playbook
- The play: no cost-per-lead number in the audience sheet without a customer count beside it. $217 looked worse than $181 and produced more customers per 1,000 leads.
- The setup: keep both kinds of audience running and pick by what is holding you back, the audiences we built for customer count and cost per customer (3.56 per 1,000 leads, $60,896 each), the platforms’ own targeting where deal size is the problem ($49,778 average won deal).
- The measurement: a close rate counted only on deals that actually closed or actually died, per audience: 19.2% on the audiences we built against 17.4% on native targeting. We could not publish retargeting’s revenue row, one advertiser held more than 40% of its closed-won deals.
Which audience produced more customers per 1,000 leads?
The audiences we built: 3.56 customers per 1,000 leads in 2025 across 104 advertisers, against 2.60 for the ad platforms’ own targeting across 72, 37% more, while costing 20% more per lead, at $217 against $181.
The scorecard is the point. Ranked on lead price, native targeting wins and a team managed on cost per lead moves budget toward it. Ranked on the customer, the order reverses: more customers per 1,000 leads, more of the decided deals won, and a lower cost per customer. Both rankings come from the same leads, over the same 2025 window, with the same rule for splitting credit. The only thing that changes is which column the report leads with. So we publish all six numbers on every audience cut, and never a cost-per-lead ranking on its own.
| Audience type | Advertisers | Cost per lead | Customers per 1,000 leads | Close rate | Cost per customer | Traceable payback | Average won deal |
|---|---|---|---|---|---|---|---|
| Audiences we built | 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 |
| Retargeting | 91 (media) | $234 LinkedIn · $120 Facebook | We could not publish the revenue row: one advertiser held more than 40% of this group’s closed-won deals. See the kill list. | ||||
Six numbers on every row is the standard table shape in this report: no audience is ever ranked on one number, and no cost number appears without a result number beside it. Green marks the better of the two on that column, orange the worse, note that they change sides between the cost-per-lead column and the customer columns. 2025 data, $29.4M of lead-generation spend, closed-won revenue with the credit split evenly across the ads that touched each deal.
Does that make the audiences we build better?
No, and this is the concession that has to ride in the same breath. Native targeting brought back more per dollar in 2025, 0.71x against 0.43x. Entirely because the deals it won were 1.9 times bigger: $49,778 on average against $26,064.
The deals were simply bigger, and we cannot say why. It could be the targeting itself, or which advertisers chose native targeting, or what they sell. Nothing in this data separates those, and we ran no test with a holdout group. So the claim stops where the evidence stops: the audiences we built turned leads into customers more often and at a lower cost per customer; native targeting sat on larger deals and therefore returned more revenue per ad dollar. Better at converting, not better at revenue. Pick by what is holding you back, build audiences for more customers at a lower cost per customer, use the platforms’ targeting where deal size is the problem.
Why does cost per lead rank audiences backwards?
Because it prices the form fill, not the buyer. In 2025 the audience with the higher lead price, $217 against $181. Produced more customers per 1,000 leads (3.56 against 2.60) and a lower cost per customer ($60,896 against $69,705).
Here is what a team actually feels. A bid or audience change that lowers the lead price usually widens who can fill in the form. A wider net lowers the price of a lead and lowers the share of those leads that ever reach a decision, so the two numbers move in opposite directions and the cheaper one looks like progress on the dashboard. The damage repeats rather than evens out, a team optimizing on lead price defunds its best-converting audience quarter after quarter and never sees the cause, because nothing in the lead report contains a customer. Automated and AI bidding makes it worse, not better: point it at a lead event inside the ad platform and it will find the widest net in the auction faster than any human could.
What about retargeting?
We can publish what its leads cost; we cannot publish what they became. Retargeting cost $234 a lead on LinkedIn and $120 on Facebook in 2025, but one advertiser held more than 40% of the group’s closed-won deals, and the rule we set in advance, that no one advertiser may swing a number, means every retargeting revenue figure stays unpublished.
That costs this page a real finding, and we would rather pay it than print the number. No retargeting close rate, cost per customer, customer count or payback figure appears anywhere in this report. If you saw one from us in an earlier draft, this rule withdrew it, and it is on the kill list with the reason. What survives is the lead-cost comparison: on LinkedIn, retargeting ran $234 a lead against $194 for audiences chasing new buyers; on Facebook, $120 against $166.
There is a second problem that would apply even with clean data. Retargeting harvests demand that other spend created, so its conversion numbers inherit the work of the campaigns that came before it. Audiences chasing new buyers, ours or the platforms’, create the demand retargeting later converts. Comparing them on conversion alone credits the harvest and ignores the planting, which is why the two belong on separate budget lines rather than in one ranking.
How could this be wrong?
Who chose which audience, and we cannot rule it out. Nobody assigned these audiences at random: different advertisers picked them, selling different products at different prices to different buyers. The gap may be pricing those differences as much as the targeting.
Four things limit that risk. The numbers are big, 104 and 72 advertisers, both far above the 8-advertiser minimum we require before publishing a revenue number, and no single advertiser holds more than 40% of either group’s spend or deals, so neither figure is one company’s result. Both use identical definitions: credit split evenly, closed-won revenue only, a close rate counted on deals that actually closed or actually died, and 12 months after the 2025 spend. The concession runs against us, not for us: the column where native targeting wins is the revenue column, and we print it in the same breath as the claim. And what we ask you to do is change a report, not buy something, put customers per 1,000 leads and cost per customer beside cost per lead, and the argument settles itself on your own data.
What should you change?
Replace cost per lead at the top of the monthly audience report with customers per 1,000 leads and cost per customer: 3.56 against 2.60, $60,896 against $69,705, across 104 and 72 advertisers. Keep lead price as a check on the work, never as the ranking.
For one quarter, fund whichever audience solves the problem you actually have, not the one with the cheaper lead. If you need more customers, the audiences you build get the budget; if you need bigger deals, native targeting does. That is where the larger won deals sat, $49,778 against $26,064, which is also why its payback reads 0.71x.
“The audiences Metadata builds deliver better ROI.” They do not, in this data: native targeting brought back 0.71x against 0.43x on deals 1.9 times bigger, and we cannot say why the deals were bigger. The claim is the cost-per-customer win, not the revenue one.
Rebuild the audience sheet this week so no cost-per-lead number appears without a customer count beside it. $217 looked worse than $181 and won 19.2% of decided deals against 17.4%. The sheet is what made the wrong audience look like the winner.
Count close rate per audience on decided deals only, the ones that actually closed or actually died, and stop pointing automated bidding at a lead event inside the ad platform. Optimizing toward the cheapest form fill widens the net, which is exactly how the $217 audience loses its budget to the $181 one.
“Retargeting is the best closer, so shift budget there.” We could not publish its revenue row, one advertiser held more than 40% of that group’s closed-won deals. Only the lead cost stands: $234 a lead on LinkedIn against $194 for audiences chasing new buyers.
What are the exact numbers?
The lead-cost rows behind the audience comparison, each paired with the result that belongs beside it. Including where that result reads “withheld”. The six-number scorecard for our audiences and the platforms’ own targeting is in the table above.
| Audience cut | Advertisers | Cost per lead | What those leads became |
|---|---|---|---|
| LinkedIn × chasing new buyers (ours or the platform’s) | 94 | $194 | LinkedIn overall: $63,312 a customer · 0.57x back, across 114 advertisers |
| LinkedIn × retargeting | 77 | $234 | withheld, one advertiser held more than 40% of the deals |
| Facebook × chasing new buyers (ours or the platform’s) | 59 | $166 | withheld, one advertiser held more than 40% of the deals |
| Facebook × retargeting | 36 | $120 | withheld, one advertiser held more than 40% of the deals |
| All advertisers, all audiences | 127 with revenue data | — | $58,887 a customer · 0.56x back · 21.2% of decided deals won |
Traceable payback = closed-won revenue we can trace back to an ad, divided by ad spend, inside 12 months. Across all 127 advertisers that was 56 cents on the dollar. Customers per 1,000 leads = the customers credited to an audience divided by its CRM-matched leads, times 1,000. Close rate = we counted only deals that actually closed or actually died. Cost per customer = ad spend divided by the customers credited to it. Average won deal = the closed-won revenue credited to an audience 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 CRM-matched leads, counted over the 12 months after the 2025 spend. The numbers we could not publish stay visible in the interactive explorer, labelled with the rule they 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. Our audiences rest on 104 advertisers, the platforms’ own targeting on 72.
- 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). Retargeting’s revenue number failed that test, so we did not publish it, not estimated, not quietly folded into the rows for audiences chasing new buyers.
- Spend floor: at least $50,000 spent behind any published number.
- What the two audience types are: the audiences we build are Metadata-built lists (by company profile, technology used, buying intent, or a named account list); native targeting is the ad platform’s own targeting options. Both chase new buyers; retargeting is scored separately.
- Scope and window: revenue numbers come from the $29.4M of lead-generation spend inside the $57.6M we analyzed in 2025 and its 154,000 CRM-matched leads, counted over the 12 months after the spend, with the credit split evenly across the ads that touched each deal and closed-won revenue only.
- What kind of evidence this is: no randomization, no holdout groups. Neither the customer gap nor the payback gap proves the ads caused the result.
- What else could explain this: who chose which audience. Different advertisers picked our audiences and the platforms’ targeting, selling different products at different deal sizes, and we cannot separate the audience from the advertiser, which is exactly why native’s deals being 1.9 times bigger is printed as an unexplained difference rather than as evidence for either side.
Full rules in the methodology, including what we refused to publish, with the exact reason each draft claim was killed.
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.
Change what the audience report leads with
The one change: rank audiences on customers per 1,000 leads and cost per customer, and move cost per lead down the page.
- Ask for your own audience table with all six numbers on it and put it next to this page’s. If cost per lead is the only column that exists, that is finding number one.
- Read the two rankings against each other. Here they disagree on the same leads: $217 against $181 a lead, 3.56 against 2.60 customers per 1,000 leads.
- Then fund the problem you actually have rather than the cheaper lead, and say which problem it is: customer count and cost per customer ($60,896) or deal size ($49,778 average won deal).
Put a customer column next to every lead column
The one change: add customers per 1,000 leads and close rate to the audience sheet before the next optimization pass, and never rank audiences without them.
- Pull leads, decided deals, won and lost by audience for the trailing two quarters. Count the close rate only on deals that actually closed or actually died, 19.2% and 17.4% are what this data reads under that count.
- Stop pointing automated bidding at a lead event inside the ad platform. Chasing the cheapest form fill widens the net, and a wider net is how the audience that produces more customers loses its budget.
- Keep both kinds of audience running and let the sheet settle it: the ones you build for customer count and cost per customer, the platform’s own targeting where deal size is the problem. Retargeting stays on its own line. It harvests demand the new-buyer campaigns created, and we could not publish its revenue row here anyway.