How to Create a Lookalike Audience on LinkedIn Ads: Setup, Minimum Sizes and Targeting Tips

If you are searching for how to build a LinkedIn lookalike audience, there is one thing you need to know before you open Campaign Manager: LinkedIn switched off its classic Lookalike Audiences feature on 29 February 2024. The button is gone. It is not hidden in a sub-menu, and no, it is not coming back.

The good news: the job that lookalikes used to do (find new people who resemble your best customers) is still very much doable on LinkedIn in 2026. It is just done with two building blocks instead of one: Matched Audiences (your data) and Predictive Audiences (LinkedIn’s AI expansion of that data).

This guide is a practical, B2B-focused walkthrough: the exact setup steps, the minimum source list sizes LinkedIn enforces, which data sources actually work, how to layer job titles on top, and what to realistically expect on audience quality and cost compared to standard attribute targeting.

The short answer

• Lookalike Audiences on LinkedIn are discontinued (sunset date: 29 February 2024).
• The direct replacement is Predictive Audiences, which uses a data source you provide plus LinkedIn’s AI to find similar members.
• Every custom audience path starts with a Matched Audience: a contact list, a company list, website retargeting, or engagement retargeting.
• The hard minimum across the board is 300 matched members before an audience can be used in a campaign.
• Expect Predictive Audiences to sit between cold attribute targeting and retargeting on both cost and quality. They are not magic, and they are not free reach.

Matched Audiences vs Predictive Audiences: what is the difference?

	Matched Audiences	Predictive Audiences
What it is	Your own data matched to LinkedIn members or companies	An AI-expanded audience modelled on a data source you choose
Closest old equivalent	Custom audiences	Lookalike audiences
Typical use	Retargeting, ABM, customer exclusion, upsell	Prospecting at scale when attribute targeting caps out
Control over targeting	High: you can narrow with job titles, seniority, function	Lower: LinkedIn decides who resembles your source
Volume	Limited by your data	Large, often hundreds of thousands of members

Minimum audience sizes on LinkedIn (the numbers that block campaigns)

This is where most B2B advertisers get stuck. LinkedIn will happily let you upload a list, then refuse to serve the campaign because the audience is too small. Here are the thresholds to plan around. This explainer is clearer than most.

Audience type	Minimum to activate	What we recommend in practice
Any campaign target audience	300 members	2,000+ for stable delivery and sane frequency
Contact list upload (emails)	300 matched members	Upload 5,000 to 10,000 rows minimum, match rates are rarely above 50%
Company list upload	300 matched companies	1,000+ companies if you plan to layer job titles on top
Website retargeting (Insight Tag)	300 members	Broad URL rules first, segment later once traffic allows
Engagement retargeting (video, Lead Gen Form, Page, event, ads)	300 members	Run a cheap video or thought leadership campaign first to fill the pool
Predictive Audience source	300 records in the source	1,000+ high quality records, ideally closed-won customers or SQLs

Important: 300 is the floor, not the target. A source list of exactly 300 will produce a weak model. If your CRM cannot give you 1,000 clean records, use website conversions or Lead Gen Form leads instead.

Step 1: Build the source (your Matched Audience)

In Campaign Manager, go to Plan > Audiences > Create audience. You will pick one of the source types below.

Option A: Website retargeting (best starting point for most B2B advertisers)

1. Confirm the LinkedIn Insight Tag is firing site-wide (check Analyze > Insight Tag for a “Verified” status).
2. Create audience > Website.
3. Name it using a convention you will still understand in six months, for example WEB_Pricing_90d.
4. Set the URL rule: “contains”, “starts with” or “exact”. Start broad (all visitors) if your traffic is under roughly 5,000 sessions per month.
5. Choose the look-back window available in your account (30, 60 or 90 days are the common options). Longer windows build size faster but dilute intent.
6. Save and wait. Population takes up to 48 hours and the audience stays greyed out until it hits 300.

Best practice: build a small hierarchy from day one: all visitors, high intent pages (pricing, demo, product), and blog readers. You will need the separation later for frequency control and for Predictive Audience sourcing.

Option B: Contact list upload (best for customer modelling)

1. Create audience > List > Contact list.
2. Download LinkedIn’s CSV template. The email column is the one that matters. You can add first name, last name, company and job title to improve matching.
3. Use work emails wherever possible. Personal Gmail addresses match poorly against LinkedIn profiles in B2B.
4. Upload up to 300,000 rows, tick the data consent box, and save.
5. Matching takes up to 48 hours. Expect a match rate of roughly 20% to 50% on a clean B2B list.

Do the maths before you upload: a 2,000 row list at a 30% match rate gives you 600 members. Usable, but thin. A 600 row list gives you 180 and will not activate.

Option C: Company list upload (best for ABM)

1. Create audience > List > Company list.
2. Fill the template with company name and, critically, company website domain. Domains match far more reliably than names.
3. Upload up to 300,000 companies. Match rates here are usually much higher, often 60% to 85%.
4. Remember: a company list targets everyone at those companies. It is almost never usable on its own, which is why job title layering matters (see step 3).

Option D: Engagement retargeting (the cheapest fuel)

LinkedIn lets you build audiences from people who engaged with your video ads, Lead Gen Forms, single image ads, Company Page, or LinkedIn events. These populate fast and cost nothing extra to build. If you are starting from zero data, run a two-week video view or thought leadership campaign purely to fill these pools, then build your Predictive Audience from the Lead Gen Form leads.

Step 2: Create the Predictive Audience (the lookalike replacement)

1. Go to Plan > Audiences > Create audience > Predictive Audience.
2. Choose your data source. You can use a contact list, Lead Gen Form leads, or conversion data captured by the Insight Tag.
3. Check that the source meets the 300 record minimum. Campaign Manager will warn you if it does not.
4. Set the audience size. Smaller means tighter resemblance to your source, larger means more reach and more dilution. Start small. The old lookalike wisdom still applies: the tightest tier almost always outperforms the broadest one on cost per lead.
5. Name it so the source is obvious, for example PRED_ClosedWon_Q3.
6. Save. Building typically takes up to 48 hours.

Which source produces the best Predictive Audience?

In order of quality, based on what we consistently see on B2B accounts:

1. Closed-won customers (highest signal, usually the smallest list)
2. Sales qualified leads or opportunities
3. High intent conversions (demo requests, pricing page conversions)
4. Lead Gen Form leads (fast to build, but includes tyre-kickers)
5. Newsletter subscribers (large but low signal, use only as a last resort)

A 1,200 record closed-won list will beat a 40,000 record newsletter list nearly every time. Quality of the seed beats size of the seed.

Step 3: Layer job titles and other filters on top

This is where LinkedIn genuinely outperforms every other platform for lookalike-style prospecting. On Meta you get a model and you accept it. On LinkedIn you can constrain the model with firmographic and professional data.

How to layer correctly

1. At campaign level, under Audience, select your saved audience under “Audiences” (Matched or Predictive).
2. Click Narrow audience further to add an AND condition.
3. Add one narrowing layer, not three. Job function + seniority, or job titles alone.
4. Watch the forecast panel on the right. If it drops below roughly 20,000 for a prospecting campaign, remove a filter.

Which filter to use with which source

Source	Recommended layer	Why
Company list (ABM)	Job function + seniority	Titles vary too much across companies, functions scale better
Website retargeting	Usually nothing, or seniority only	The visit is already the qualifier, extra filters just starve delivery
Predictive Audience	Company size and industry, plus seniority if reach allows	Keeps the model inside your actual ICP band
Contact list	Exclusions rather than narrowing	Exclude existing customers so you do not pay to reach them

A practical note: Campaign Manager does not always let you narrow a Predictive Audience as aggressively as a Matched Audience. If the option is unavailable or greyed out, build the filter into the source instead. In other words, if you only want VPs at 500+ employee companies, feed the model a source list that only contains VPs at 500+ employee companies.

Always keep these three exclusions active on prospecting campaigns:

• Existing customers (contact list exclusion)
• Current open opportunities (so sales conversations are not undercut by generic top-funnel ads)
• Your own employees (company exclusion)

Audience quality and cost: realistic expectations

Let us be blunt about what these audiences deliver. LinkedIn is expensive and no audience type changes that. What changes is efficiency further down the funnel.

Audience type	Typical CPC range (B2B, EN markets)	Scale	Lead quality
Website retargeting	$6 to $12	Low	Highest
Contact list / ABM company list	$10 to $20	Low to medium	High
Predictive Audience	$8 to $16	High	Medium to high
Standard attribute targeting	$8 to $15	Very high	Variable

Treat these as planning ranges, not guarantees. Market, seniority and ad format move them significantly, and Sponsored Messaging or Conversation Ads follow different economics entirely. The point is broken down further in this analysis.

What we typically observe: Predictive Audiences deliver a similar or slightly higher CPC than attribute targeting, but a lower cost per qualified lead, because the model filters out people who technically match your job title criteria yet never buy. The gain shows up in MQL to SQL conversion, not in the CPC column. If you only judge the test on CPC, you will kill a winner.

Common mistakes to avoid

• Seeding the model with your newsletter list. Big list, weak signal, mediocre audience.
• Over-layering. Predictive Audience + job title + seniority + company size + years of experience equals no delivery and a wasted budget cycle.
• Judging after 4 days. LinkedIn’s learning phase plus a B2B sales cycle means you need at least 3 to 4 weeks and roughly 50 conversions before you draw conclusions.
• Forgetting exclusions. Paying $14 per click to reach current customers is a silent budget leak.
• Using stale lists. A contact list from three years ago matches badly and models badly. Refresh source lists quarterly.
• Ignoring the 300 floor when segmenting. Splitting one 900-member audience into four segments gives you four audiences that cannot run.

A 30-day test plan you can copy

1. Days 1 to 3: Verify the Insight Tag, build website audiences (all visitors 90 days, high intent 90 days), upload a customer contact list and a suppression list.
2. Days 3 to 5: Once lists are matched, create one Predictive Audience from your best source (closed-won or SQLs), tightest size setting available.
3. Days 5 to 30: Run three campaigns with the same creative and offer: (A) standard attribute targeting as your control, (B) Predictive Audience with a light firmographic layer, (C) website retargeting for the bottom of funnel.
4. Budget split: roughly 40% control, 40% Predictive, 20% retargeting.
5. Measure: cost per lead and, more importantly, cost per qualified lead pulled from your CRM, not from Campaign Manager.
6. Then scale: shift budget to the winner and build a second Predictive Audience from a different source to test source quality against source quality.

FAQ

Does LinkedIn still have lookalike audiences?

No. LinkedIn discontinued Lookalike Audiences on 29 February 2024. Existing lookalike audiences stopped serving and the creation option was removed from Campaign Manager. Predictive Audiences are the intended replacement.

What are LinkedIn Matched Audiences?

Matched Audiences is LinkedIn’s custom audience system. It lets you target members based on your own first-party data: website visitors captured by the Insight Tag, uploaded contact lists, uploaded company lists, CRM integrations, and people who engaged with your ads, Page, events or Lead Gen Forms.

How does a lookalike audience work?

A lookalike model takes a source list of people you already value, analyses shared attributes and behaviour patterns, then finds other platform members who resemble that profile. On LinkedIn today, that logic lives inside Predictive Audiences, which additionally draws on professional data such as job function, seniority, industry and company size. It is argued more carefully on workshopdigital.com.

What is the minimum audience size for LinkedIn Ads?

300 members. Any audience below that threshold cannot be used to run a campaign. For Predictive Audiences the data source also needs at least 300 records, and we strongly recommend 1,000 or more for a usable model.

How long does a LinkedIn audience take to build?

Up to 48 hours for list uploads, website audiences and Predictive Audiences. Plan your launch date accordingly rather than uploading a list the morning a campaign is due to go live.

Can I add job titles on top of a Predictive Audience?

Sometimes, depending on the campaign objective and what Campaign Manager allows at the time. When narrowing is restricted, the reliable workaround is to bake the filter into the source list so the model is trained only on the roles you care about.

What is the 4-1-1 rule on LinkedIn?

It is an organic content guideline, not an ads setting: for every one self-promotional post and one shared post from someone else, publish four pieces of genuinely useful content. It is a good discipline for the Page content that feeds your engagement retargeting audiences, which in turn feed your paid funnel.

Final word

The disappearance of lookalikes made LinkedIn targeting more demanding, not less effective. The advertisers winning today are the ones with clean first-party data, a disciplined audience naming structure, and the patience to judge results on pipeline instead of CPC.

If you want help building the source data, the audience architecture and the creative that makes it all work, the team at King Content Agency runs this playbook for B2B advertisers every week. Get in touch and we will audit your current LinkedIn setup.