A Lookalike Audience is a Meta- or Google-built segment of users who share statistical traits with a source audience you provide — usually your existing customers, your highest-LTV cohort, or visitors who took a specific action. The platform's model finds people who 'look like' your source.
How they're built
You upload a customer list (hashed), point at a website-pixel event (e.g. all purchasers in the last 90 days), or use a Meta event from your pixel. Meta's model finds users in the country you specify whose behaviour patterns align with your source. You pick a percentage — 1% is the closest match, 10% is the broadest.
Why they matter
Lookalikes are the most reliable cold audience for direct-response advertisers. Broad targeting works for big brands with broad appeal; lookalikes work for niche products because they encode 'who already buys this' into the targeting, even after iOS 14.5 weakened other signals.
How to build a good lookalike
- Source quality matters more than source size. 500 high-LTV customers beats 50,000 trial signups.
- Use first-party data: customer list with email + phone + name. Pixel-derived lookalikes are weaker post-iOS.
- Start at 1–2% lookalike for top-performing source, then test 3–5% as a separate ad set
- Refresh quarterly — your customer base shifts and lookalikes should follow
- A 'value-based' lookalike (where the source carries an LTV column) outperforms a flat lookalike