What is a lookalike audience and how is it built?

A lookalike audience (Google calls its equivalent similar segments) is a group of people an ad platform assembles because they resemble an existing group you supplied — your customers, your website visitors, or your highest-value users.

How it works:

You provide a source audience, either uploaded (hashed emails or phone numbers) or collected via a pixel. The platform matches those people to its own user profiles.

It then examines what that group has in common across the enormous number of signals it holds — demographics, interests, behaviours, device patterns, engagement history, purchase signals. Crucially, you never see which attributes it found, and the model is not built from the characteristics you would guess.

It identifies other users scoring highly on similarity and makes them targetable.

Audience size is the main lever. A 1% lookalike is the closest match and the smallest pool; expanding to 5% or 10% reaches far more people who resemble your customers less. Tighter is usually better for conversion campaigns, broader for awareness or when the tight audience exhausts.

What determines whether it works:

Source quality dominates. A lookalike built from all website visitors models people who visited a website. One built from repeat purchasers models people who buy repeatedly. The second is almost always better, and the difference is larger than most other optimisations available to you.

Source size matters — platforms typically want at least a thousand matched people, and quality still beats quantity.

Recency matters. Behaviour drifts; a source list from three years ago models a different market.

Two caveats. Privacy regulation and platform changes have reduced matching rates and constrained the feature in some markets. And modern broad-targeting algorithms increasingly find the same people without an explicit lookalike, so the technique matters less than it did.