A lookalike audience lets you say "find more people like these" without knowing what those people have in common. You give Meta a source, such as your customers, site visitors or engaged followers, and Meta finds people in your chosen country who share patterns with them.
How lookalikes are built
- Pick a source. Usually a custom audience: a customer list, pixel purchasers or people who engaged with your Instagram account. Meta sets a minimum source size (at the time of writing, at least 100 people from a single country), but a larger, higher-quality source gives it a clearer pattern.
- Choose a location. The lookalike is built within the countries or regions you select.
- Choose a size. From 1% to 10% of the location's population. 1% is the closest match. Larger percentages trade similarity for reach.
Meta refreshes lookalikes periodically while you use them, so one built on a live pixel audience stays current. If your source is an uploaded list, refresh the list so the seed reflects today's customers.
Value-based lookalikes
If you upload a customer list with a value column, such as each customer's total spend, Meta can weight the lookalike towards people who resemble your highest-value buyers. That usually makes a better seed than "all customers", because it points Meta at people likely to have a higher customer lifetime value, not just a first order.
Their role now that broad targeting often wins
Meta's delivery system has become good at finding buyers through broad targeting, and in many accounts a lookalike no longer beats a wide-open ad set. In Advantage+ setups, a lookalike usually acts as an audience suggestion: Meta starts there but may deliver beyond it when it predicts better results (this used to be called lookalike expansion).
Lookalikes still earn their place when:
- Your pixel is new and Meta has little purchase data to learn from.
- Your best customers came from outside Meta, such as offline sales, marketplaces or a CRM, so the pixel never saw them.
- The product is niche and broad delivery keeps finding the wrong people.
How to test a lookalike
Run it against a broad ad set with the same creative and budget, ideally for a week or more. Judge on cost per purchase and the share of new customers in your store data, not click-through rate. If the lookalike doesn't clearly win, keep it as a suggestion inside your broad setup or an Advantage+ sales campaign rather than giving it an ad set of its own.
Common mistakes
- Weak seeds. A list of newsletter sign-ups tells Meta less than a list of repeat buyers. When building a seed, quality beats size.
- Many lookalikes in separate ad sets. They overlap heavily and compete in the same auctions.
- Forgetting consent. Customer lists must be collected and used lawfully. Hashing alone doesn't make an upload compliant.
The D2C playbook shows where lookalikes fit alongside broad and retargeting campaigns.
Frequently asked questions
What size lookalike audience should I use?
Start with 1% if you want the closest match, or a 1–3% range if you need more reach. Larger percentages add people who are less similar to your source. Before building many sizes, test one lookalike against broad targeting.
Do lookalike audiences still work on Meta?
They still work, but they matter less than they used to. Meta's broad and Advantage+ audience delivery often matches or beats them, especially in accounts with plenty of purchase data. Lookalikes are most useful for new pixels, offline customer data and niche products.
How many people do I need to create a lookalike audience?
Meta sets a minimum source size, at least 100 people from a single country at the time of writing, but that's a floor, not a target. A larger, higher-quality source such as repeat buyers gives Meta a clearer pattern to match. Check Ads Manager for the current requirement.
