Do LinkedIn ads work? A decision table

LinkedIn ads work when 1 customer is worth more than LinkedIn charges to win one, and in Metadata's 2025 data from 153 advertisers that cost was 31,000 USD for a customer at a small company and 78,000 to 103,000 USD at enterprise and mid-market companies. They also need an audience of at least 300 members, which is the smallest LinkedIn will serve, plus a budget that reaches each member 1 to 8 times a month and a measurement window longer than the sales cycle.

Updated 2026-11-07. Every figure names its source and the year of its data.

Do LinkedIn ads work? The decision table#

Put your own numbers into each row and read across to the column they land in.

Any row in the Fix first column is the thing to fix before you spend more.

The month bands follow annual contracts. At 12 months or fewer, a customer on a 1-year contract earns back what it cost to win before it comes up for renewal. Past 24, it has to renew twice before LinkedIn's share is back.

If an opportunity in the CRM can't be traced back to the ad set that started it, the last row stays in Fix first no matter how the other 6 look.

What public data says about LinkedIn ad returns#

5 public reports put a number on LinkedIn ad performance. No 2 of them count the same thing, so the table says what each one measures.

Dreamdata and Metadata both compare channels, and LinkedIn comes out ahead in each: above Google Search and Meta in one, above the all-channel average in the other. The 2 that measure time explain why a short test misses most of it. HockeyStack counted 24 days from lead to sales-qualified opportunity, and Dreamdata's buyer journey stretched to 272 days in its 2026 report, up from 211.

Why the return figures disagree#

Metadata's 10.20 USD is sourced pipeline, the opportunities lead-gen ads started, divided by that spend. It comes from the 90 of Metadata's 153 advertisers that connected a CRM, a smaller group than the one behind its cost per lead. ZenABM's 5.21 USD is a looser credit from a different kind of program: any pipeline at a target account that saw 3 or more impressions from an account-based campaign. Dreamdata's 121% counts closed revenue instead, deals won within 12 months, each split across every touch by a data-driven model.

Keep in mind who published them. All 3 sell measurement or account-based tools, and Dreamdata describes itself as a LinkedIn Marketing Partner. Metadata's 153 advertisers ran their paid media through Metadata's product and Dreamdata's data comes from its own customers, so both samples are companies already paying to trace pipeline back to ads.

The figures also clash with pages that rank for this question, and with each other:

A worked payback example#

The formula is payback months = cost per customer / monthly gross margin per customer. Below, the contract and margin are made-up example inputs, and the cost per customer is Metadata's 2025 LinkedIn figure, split by the size of the company that bought.

Take a 40,000 USD a year contract at 80% gross margin. That's 32,000 USD of margin a year, or 2,667 USD a month.

The odd one is mid-market, which cost 25,000 USD more per customer than enterprise. Metadata's note on that chart is that 1 blended number hides a 3x gap, so use the row for the companies you actually sell to.

Coming at it from the spend side gives a similar answer. Spend 5,000 USD a month for 12 months, 60,000 USD in all, and Dreamdata's 121% return on ad spend turns it into 72,600 USD of attributed revenue. At 80% margin that's 58,080 USD, about 97% of the spend back by month 12, close to the small-company row.

Margin and contract size are the inputs you can move. Raising margin to 85% only takes the small-company row from 11.6 to 10.9 months. A 60,000 USD contract does far more: 7.8 months at small companies, and enterprise falls to 19.5, which is Borderline instead of Fix first.

Retargeting is the other lever, with thinner evidence. Metadata found it sourced 9.6 USD of pipeline per 1 USD at 196 USD a lead, slightly more per lead than cold prospecting at 187 USD, and its advice is to fund retargeting on pipeline per dollar. The report doesn't say whether that figure covers LinkedIn alone.

When LinkedIn ads don't work#

6 setups make LinkedIn look like it failed, and 5 of them can be fixed inside the account. The last one, a price point that's too low, is a good reason to put the budget somewhere else.

The audience can't clear 300 members. LinkedIn won't serve an ad set under 300 (LinkedIn Help a423690). In the audience sizing skill's made-up example, a 1:1 audience at 1 logistics company fell to 260 when it was limited to vice presidents and C-level titles, and only cleared the floor, at 610, once directors went back in.

The budget sits at the 10 USD floor on a large audience. 10 USD a day for 30 days buys about 4,800 impressions at ZenABM's 62.67 USD US median per 1,000. On a 50,000-member audience that's under 0.1 impressions per member a month, and the auction decides which few people see anything.

The verdict comes inside 30 days. Dreamdata's 2026 report puts the business buyer journey at 272 days, and Ehrenberg-Bass Institute research with John Dawes says about 5% of business buyers are in market at a time. A 30-day read can only see the few who were already shopping, while most of the people the ads reached won't be ready for months.

The ads are judged on click-through rate. ZenABM found a slightly negative correlation, -0.170, between click-through rate and pipeline in 2026, and the top quarter by pipeline per dollar in its 2025 account-based data ran a 0.21% median click-through rate against 0.69% for the typical company. Click-through rate still helps you spot ads nobody reads, which is how the click-through rate guide uses it, but it can't tell you whether the channel pays.

The price point is too low. A 3,000 USD a year product at 80% margin earns 200 USD a month, so Metadata's 31,000 USD small-company cost per customer takes about 13 years to come back. Social Surge Marketing, an agency, starts its pros and cons page (2025-06-06) by saying "we don't recommend LinkedIn Ads very much anymore", and puts it down to opportunity cost.

Audience expansion spreads spend past the list. LinkedIn turns it on by default where it's available and adds members "with similar attributes", who aren't counted in the forecast (LinkedIn Help a418929). On an account list, that means paying for people at companies you never picked. Switch it off for account-based and retargeting ad sets, or create them through Adwarp, which leaves it off.

What Adwarp's rules stopped#

Adwarp was built while running LinkedIn campaigns through it, and 2 of these patterns became rules. Every audience gets sized before an ad set exists: 169 audience estimates so far, each checked against the 300-member floor, so an ad set that couldn't serve never got built. Every LinkedIn ad set Adwarp creates also has Audience Expansion and Audience Network turned off, and no Adwarp tool turns them back on (safety).

A third rule runs on Mondays and leans on click-through rate, which might look odd after the ZenABM finding. The linkedin-ads-pause-rules skill pauses an ad that has passed 1,500 impressions under the floor for its format: 0.29% for single image, 0.22% for carousel and 0.17% for video, each 70% of ZenABM's 2025 median for that format. It pauses at most 5 ads a run, protects the top 5 by conversions and the last active ad in each ad set, and shows a dry-run table before anything changes. All it clears out is single ads nobody engages with. Whether LinkedIn works at all gets answered in the CRM.

Every change also lands in the change journal, 1,125 since June, including 29 budget changes, 23 bid changes, 16 rotation switches and 36 notes on why. When results drop, adwarp changelog list shows what changed and when, and adwarp lift compares the 14 days before a change with the 14 after.

Pros and cons, with numbers#

These are medians and averages across many advertisers, roughly what anyone pays. What differs between 2 companies is what 1 customer is worth to them, and that decides whether 8.99 USD a click is a lot. The full price breakdown is in LinkedIn Ads cost.

How to test LinkedIn ads in 90 days#

90 days is enough to see whether the ads reach the right accounts and start pipeline. These steps set the test up so the read at day 90 means something.

1. Size the audience. Narrow by job function, then seniority, then company filters, and check the estimate after each step against 300 and the range for the goal. The linkedin-audience-sizing skill walks it with estimate_audience. More on targeting at LinkedIn audiences.
2. Set the budget from the frequency band. Monthly budget = audience x impressions per member x cost per 1,000 / 1,000. Aim for 1 to 8 impressions per member a month, using your own cost per 1,000 when you have one.
3. Run 1 lead form ad set and 1 retargeting ad set. Lead forms were the cheaper lead in Metadata's data, and retargeting is where Metadata says to judge spend on pipeline per dollar.
4. Log the hypothesis before launch. Write down what should happen, the metric and the read date. The ad-experiment-lift skill logs it with log_ad_change and reads before and after lift with compute_lift on equal windows.
5. Judge each ad after about 1,000 impressions or 3 conversions. The linkedin-ads-performance-review skill holds every call to that floor. At 10 USD a day and a 62.67 USD cost per 1,000, an ad set buys about 160 impressions a day, so 3 ads take about 19 days to reach 1,000 each.
6. Read pipeline in the CRM at day 90, and again at the length of your sales cycle. Day 90 shows whether the right accounts are turning into opportunities. Whether they pay back takes the second read.

Steps 1, 2 and 4 look like this from the terminal, using the made-up audience in the sizing skill's examples:

At 1.6 impressions per member a month, the budget reaches the audience without wearing it out. Run the same check on a 1,000-member account list at that budget and it comes back at 86 impressions per member, far past 8, so either the budget drops or the list grows.

By Stan Rymkiewicz, maintainer of adwarp. Published November 7, 2026.