Lookalike Audiences

A lookalike audience starts from a list of your best customers. The ad platform then finds new people who share their traits. The quality of that seed list matters more than any other setting.

How it works

You give the platform a seed audience: a customer list, a pixel audience, or an engagement audience. The platform looks for patterns across that group, including demographics and behavior. It then ranks its users by how closely they match and builds an audience from the top matches.

A smaller percentage means a closer match and less reach. A larger percentage means more reach and a looser match.

On Meta, lookalikes still exist, but with Advantage+ audience they often act as a suggestion. The system can deliver beyond them when it predicts better results. Google retired Similar Audiences in 2023. On Google, you upload customer lists as audience signals and let optimized targeting expand from there.

How to do it

1. Build a seed from your best customers: top customers by lifetime value, repeat buyers, or high order value buyers.
2. Avoid “all customers” as a seed. One-time discount buyers and frequent refunders weaken the signal.
3. Upload customer lists with email, phone, and name to get the best match rate.
4. Pass customer value with the list where the platform supports value-based seeds.
5. Check the match rate. A low rate means your data needs cleaning before the lookalike is worth testing.
6. Create lookalikes at a few size tiers from the same seed, for example 1%, 1-3%, and 3-5%.
7. Launch your best creative against the smallest tier first. Expand to larger tiers when it stops scaling.
8. Exclude current customers and recent site visitors from every lookalike campaign.
9. Rebuild seeds every few months with fresh transaction data. Compare the new version against the old one.

What to measure

• Seed match rate: share of uploaded records the platform matched to user profiles.
• CPA by tier: cost per acquisition for each lookalike size, compared side by side.
• Lookalike vs. broad CPA: lookalike campaign CPA compared to a broad or Advantage+ audience campaign with the same creative.

Best practices

• Use value-based seeds when you can. They tell the platform which customers are worth the most.
• Build lookalikes from different seed types (buyers, high-value customers, engaged subscribers) and test them against each other.
• Seed composition often matters more than the percentage tier.
• Test lookalikes against broad targeting. On many accounts, broad now performs as well or better.

Common pitfalls

• Using a seed that is too small. The platform has too few examples to find a real pattern.
• Never refreshing seeds. An old customer list describes who your customers were, not who they are.
• Stacking interest targeting on top of a lookalike. It shrinks the pool and limits the algorithm.
• Assuming a lookalike is a hard boundary on Meta. With Advantage+ audience, delivery can go beyond it.