Lookalike audience
Also: Similar audience, LAL, Lookalike modeling, Audience similaire, Audience de jumeaux
An audience created by ad platforms to target new prospects who resemble your best existing customers, based on shared traits and behaviors.
What it is
A lookalike audience (also called a similar audience) is a targeting segment built by an advertising platform. You provide a source audience (a "seed") of people you already value, for example past purchasers or high lifetime value customers. The platform then analyzes the shared traits of that seed and finds new users in its network who statistically resemble them.
The key idea: you are not choosing targeting rules by hand (age, city, interests). Instead, the platform infers the pattern from your seed and expands it to strangers who look similar.
Why it matters
- Prospecting at scale: it moves you beyond retargeting (people who already know you) into finding net new prospects.
- Efficiency: a good seed usually beats broad demographic targeting on cost per acquisition.
- Quality control: the better your seed, the better the lookalike. Feeding "all buyers" produces a weaker result than feeding "top 10 percent by revenue."
How it is used in practice
1. Build the seed from a customer list (emails, phone numbers, hashed identifiers) or from platform events (purchases, sign ups).
2. Upload or select the seed inside the ad platform.
3. Choose the expansion size, often as a percentage of the target country's population. A 1 percent lookalike is tighter and more similar; a 5 to 10 percent lookalike is broader and less precise.
4. Launch campaigns against the resulting audience and measure incremental conversions.
Common pitfalls:
- Seeds that are too small (aim for at least a few thousand quality records).
- Stale seeds that no longer reflect current best customers.
- Privacy and consent gaps when uploading personal data.
Concrete worked example
A subscription software company exports its top 5,000 customers by annual contract value. It hashes their emails and uploads them as a seed. The platform builds a 1 percent lookalike in the target market, roughly 250,000 people.
- Before: broad interest targeting delivered a cost per trial of 40 dollars.
- After: the 1 percent lookalike delivered a cost per trial of 24 dollars with higher trial to paid conversion.
The team then tests a 2 percent version to grow reach once the 1 percent audience saturates, accepting a slightly higher cost for more volume.
See also
Frequently asked questions
What is a lookalike audience in advertising?
A lookalike audience is a targeting segment built by an ad platform from a source audience you supply, called a seed. The platform analyzes what your seed members have in common and finds new users in its network who statistically resemble them. You don't write the targeting rules yourself: the pattern is inferred from your data and extended to strangers.
What is the difference between a lookalike audience and retargeting?
Retargeting addresses people who already interacted with you: visitors, cart abandoners, existing customers. A lookalike audience targets people who have never heard of you but resemble your best customers. Retargeting harvests existing demand, lookalike audiences do prospecting at scale.
How large should the seed audience be for a lookalike to work?
Aim for at least a few thousand quality records. Seeds that are too small give the platform too little signal to find a reliable pattern. Freshness matters as much as size: a seed that no longer reflects your current best customers will produce a lookalike aimed at yesterday's buyers.
Should I choose a 1 percent or a 10 percent lookalike?
The expansion size is usually expressed as a percentage of the target country's population. A 1 percent lookalike is tighter and closer to your seed; a 5 to 10 percent lookalike gives far more reach with less precision. Start narrow to validate performance, then widen once the tighter audience saturates and you need volume.
Does a lookalike audience actually lower cost per acquisition?
It usually does when the seed is well chosen. In one documented case, a subscription software company uploaded its top 5,000 customers by annual contract value as a hashed seed, built a 1 percent lookalike of roughly 250,000 people, and saw cost per trial drop from 40 to 24 dollars with better trial to paid conversion. Feeding "all buyers" instead of "top 10 percent by revenue" typically produces a weaker result.