Lookalike audience
A lookalike audience is a targeting segment that an advertising platform builds by analysing a source list of existing customers or engaged users — the seed — and finding new people whose attributes and behaviour most closely resemble them.
The mechanism is the same everywhere even though the names differ. You supply a seed: a customer list, a website or app event audience, a video-viewer segment, or people who engaged with your page. The platform models what those people have in common, ranks its wider user base by similarity, and serves you the top slice. On Meta the slice is expressed as a percentage of a chosen country's population — 1% is the tightest and most similar, 10% the broadest and most diluted. Google's closest equivalents are Similar Segments and Customer Match expansion; LinkedIn offers lookalike audiences built from matched or website audiences plus audience expansion; TikTok and Microsoft Advertising have their own variants of the same idea.
Seed quality decides everything. A seed of a few thousand genuinely high-value customers outperforms one of a hundred thousand undifferentiated site visitors, because the model learns whatever pattern you feed it. Two habits pay off: seed from your best customers rather than all customers — top-quartile lifetime value, repeat purchasers, closed-won deals rather than raw leads — and keep the seed reasonably fresh so it reflects current demand rather than last year's.
Sizing is a trade-off, not a dial to maximise. Tighter percentages give higher precision and higher CPMs on a smaller pool; broader ones give reach and cheaper impressions but drift towards generic targeting. Many advertisers run a 1-2% and a 5-6% variant as separate ad sets and let performance decide. Also worth knowing: lookalikes are built within a single country or region, they need enough matched seed records to build at all, and privacy rules — consent, hashing of uploaded data, and a lawful basis under regimes such as GDPR and Turkey's KVKK — apply to any customer list you upload.
Opus Growth lets you build and manage these audiences conversationally across the platforms we support, including website and lookalike audience creation, interest search and reach estimation. Because audience creation is a write action, it runs as a dry run first — you see the exact seed, region and size before anything is created, and nothing reaches the platform without your approval.
Frequently asked questions
Platforms typically need at least a thousand or so matched records, but quality matters more than raw size. A tightly defined seed of a few thousand high-value customers usually outperforms a large, undifferentiated list.
Not automatically. 1% is the most similar and usually converts best per impression but has limited reach and higher CPMs; 10% reaches far more people with weaker resemblance. Testing two sizes as separate ad sets is the standard approach.
Similar segments and Customer Match expansion play the same role — Google models your uploaded or first-party audience and targets users with comparable signals, though the controls and naming differ from Meta's percentage-based sizing.
