The blog · Demand Generation · From The Demand Compass · 7 min
What is the ROAS of LinkedIn ads vs Google Search vs Meta for B2B?
LinkedIn returned 121 percent on ad spend in the 2026 benchmark, Google Search 67 and Meta 51, while budgets stayed where they were. The gap between the two tables is the trade sitting in the open.
In Dreamdata's 2026 LinkedIn Ads Benchmarks Report, built on 66 million sessions across 3.5 million complete B2B journeys, LinkedIn ads returned 121 percent on ad spend, Google Search returned 67 percent, down from 78 the year before, and Meta returned 51 percent. One channel in the table gives back more than it takes in. Budgets have not followed: Google still holds 46 percent of B2B ad spend against LinkedIn's 41 and Meta's 8, roughly the same allocation as the year before, while the gap between the top and bottom of the table widened. The distance between the two tables is a reallocation sitting in the open.
What is the ROAS of each channel for B2B?
Dreamdata's 2026 benchmark puts LinkedIn at 121 percent return on ad spend, Google Search at 67 percent and Meta at 51 percent, measured across thousands of B2B companies and 3.5 million complete customer journeys. LinkedIn is the only channel above water. Google Search fell from 78 percent the year before. Meta nearly doubled, from 29 to 51 percent, the biggest improvement of any platform, and it is still under water. Improvement and viability are different columns now. Source: Dreamdata, LinkedIn Ads Benchmarks Report, 2026
Jared Robin, who convenes tens of thousands of revenue professionals at RevGenius, wrote the field note in our book that these figures come from, and his reading is the one we share: this is a table most CMOs will not want forwarded to their CFO, and it will be forwarded anyway.
| Channel | ROAS, 2026 | Change | Share of B2B ad budget |
|---|---|---|---|
| LinkedIn ads | 121% | Up across two report years | 41% |
| Google Search | 67% | Down from 78% | 46% |
| Meta | 51% | Up from 29% | 8% |
Marketing is about to be graded the way a portfolio is graded: by return, by line, in public.
Why did the unit economics invert?
LinkedIn's clicks got more expensive while its accounts got cheaper, and a metric built on counting clicks reads that backwards. Cost per click on LinkedIn rose 11.8 percent to €5.98, among the most expensive clicks in advertising, while cost per company influenced fell 54 percent to €70. A regime built on counting cheap actions reads that as deterioration. A regime built on revenue reads it as the whole point: expensive clicks, cheap accounts. Source: Dreamdata, LinkedIn Ads Benchmarks Report, 2026
Expensive clicks, cheap accounts.
That is the difference between grading a channel by hand-raisers and grading it by return. An MQL target could be hit by any channel that produced form fills, including cheap and useless ones. A return table exposes every channel at line-item level. Google Search holding 46 percent of B2B budgets while returning 67 cents on the dollar is exactly the kind of line that used to hide inside a healthy lead count, and it cannot hide in the new table.
Why haven't budgets moved to match?
Because allocation moves on habit, agency incentive and last year's plan, while returns move on buyer behavior. The Google network still commands 46 percent of B2B ad spend against LinkedIn's 41 and Meta's 8, essentially the same split as the year before, while the return gap widened. The distance between those two tables is the reallocation trade currently sitting in the open, and the first team in a category to make it captures the arbitrage while competitors are still defending their MQL targets.
The people carrying revenue targets have already moved. Inc. surveyed the CEOs of the Inc. 5000 and 57 percent named LinkedIn the most important platform for growth, a majority every year since 2020. LinkedIn now takes 41 percent of B2B paid social budgets. When the CEO reads one feed and one attribution table and both point at the same channel, the marketing team's dashboard debate has been settled above their heads. Source: Inc., Inc. 5000 CEO survey
What is behind the decline in search?
Buyers moved their research somewhere ads do not reach and forms do not gate. Demandbase measured ChatGPT referrals to B2B websites nearly quadrupling in a year, from 645,000 monthly visits to 2.6 million. Dreamdata's journey data agrees: the average B2B buying journey now runs 272 days, up from 211, and 81 percent of it happens before sales ever sees a name. Form-to-SQL is down to 92 days and SQL-to-close to 52. The stages a lead count measures are the shrinking end of a lengthening journey. Source: Demandbase platform data, 2026; Dreamdata, LinkedIn Ads Benchmarks Report, 2026
The dark channel, the communities, Slack threads, podcasts, AI assistants and LinkedIn feeds where the research runs, cannot be gated. The only move available is to already be present when the research starts. And increasingly the research is done by machines: the majority of AI assistant answers cite editorial or earned media rather than brand-owned pages, which makes press coverage and community reputation inputs to whether an assistant mentions you at all. That is a brand and community investment on a 272-day fuse, and it is precisely the spending a lead-count regime always starved, because it could never claim credit inside a quarter.
What is the caveat on these numbers?
The caveat belongs in the open: Dreamdata sells attribution software, and its own report flags that this year's numbers are not like-for-like. A new IP-resolution engine identifies up to 15 times more companies than the previous year's, and a LinkedIn integration made organic impressions measurable for the first time. Treat 121 percent as one vendor's lens, not a law of nature. The direction, though, is corroborated from outside the vendor's data: LinkedIn's return rose across two report years while Google's fell, buyer research visibly moved into AI assistants, and the CEOs voted with their own posting habits. You can argue with the precision. Arguing with the direction requires explaining a lot of coincidence.
What should a marketing leader do with the table?
Three moves, in the order they will be forced on you anyway. First, run the table on yourself before finance does: pull twelve months of spend and closed revenue by channel and compute the ROAS line your CFO will eventually compute regardless. If your search line is under water, being the person who found it is a very different meeting from being the person it was found on.
Second, change compensation before dashboards. A demand generation team paid on MQL volume will manufacture MQL volume; that is what compensation plans are for. Demand Gen Report's 2026 benchmark survey found the leading teams have already moved comp and reporting to sourced revenue and influenced pipeline. If the comp plan still says leads, the dashboard migration is decoration on top of the old behavior. Source: Demand Gen Report, 2026 Demand Generation Benchmark Survey
Third, treat the dark 81 percent as territory rather than noise. The research phase happens where a form cannot follow, so the work is to be present when it starts and to measure presence at the account level, which is what the awareness axis of the Demand Compass exists to do.
Related questions
Does a 121 percent ROAS mean LinkedIn ads are profitable for every B2B company?
No. It is a benchmark across thousands of companies through one vendor's attribution. Your number depends on your deal size, your market and how well your ads land on accounts that already know you. Compute your own line before moving budget.
Should we cut Google Search entirely?
Not from this table alone. A 67 percent return across the benchmark hides companies well above and below it. Run your own twelve-month channel table, then move budget from the lines under water toward the ones above it, in steps you can measure.
Why does Meta's improvement not matter if it is still under water?
Because improvement and viability are different columns. Nearly doubling from 29 to 51 percent is real progress; returning 51 cents on the dollar is still a loss. The trend is worth watching, the allocation is not yet worth making.
How do these figures relate to the MQL?
They are the reason the MQL is being replaced. A lead count was a proxy that survived while the real number was unmeasurable. Once finance can read return by channel, nobody needs to know how many people filled in a form.
