Lookalike audience
A lookalike audience is a new group of people who resemble customers you already have. You give the ads tool a list of buyers or email subscribers, and it finds other people who look similar. Those new people have never visited or bought from you. Google calls its version similar segments (the same idea: new people who resemble your list).
Theodor Lindfors, Founding Marketer ·
What a lookalike audience is
You start with a seed: a custom audience of buyers (a Meta list of people who already interacted with you), a Customer Match list (a hashed customer file uploaded to Google Ads), video viewers, or app users. The platform models shared traits and behaviour, then finds users who match that pattern.
The output is a cold audience. Nobody in a lookalike has heard of you, which puts it squarely in prospecting (ads to people who have never engaged) even though it was built from your best customers.
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand advertises on Meta (Facebook and Instagram), TikTok, and YouTube. The team exports 4,000 people who bought at least twice. That file of 4,000 repeat buyers becomes the seed. Meta builds a 1% United States lookalike, a few million people who look like those buyers on paper, and none of them are on the original list of 4,000.
Why lookalikes matter
Lookalike audiences are the shortest route from first-party data to new reach. Instead of guessing interests, you hand over the people who already paid and let the auction find more of them.
Lookalikes have also lost ground. Broad targeting plus strong creative now beats a stack of narrow lookalikes on Meta (Facebook and Instagram) more often than it used to, and Advantage+ products (Meta's automated ad tools) lean on the same modelling under the hood.
How to read lookalike performance
Judge the seed before the audience. A 100-person seed of trial signups and a 20,000-person seed of repeat buyers produce very different models from identical settings.
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand spends $20,000 in a month on the 1% buyer lookalike. Meta (Facebook and Instagram) credits 200 purchases. Divide $20,000 of lookalike spend by 200 purchases. That is $100 per purchase. The pack sells for $40, so this only works if those purchases are new customers with repeat value attached. The drink brand checks: 170 of the 200 purchases had never bought before. New-customer cost is $20,000 of lookalike spend divided by 170 new customers, about $118 per new customer. A second ad set seeded on all site visitors, not buyers, spent the same $20,000 and brought 90 new customers. The seed was the difference, not the lookalike percentage.
Refresh the seed on a schedule. A list built two years ago models the customer you had then. Also check overlap: five lookalikes from similar seeds mostly bid against each other.
Common lookalike mistakes
- Seeding on all site visitors, which models browsers instead of buyers.
- Running many narrow lookalikes at once and splitting the same pool.
- Never refreshing the source list as the customer base changes.
Lemonado
How Lemonado helps with lookalike audiences
A lookalike is only as good as the seed list. Lemonado reads customer and revenue data next to ad performance, so you can see which audiences were built from people who paid.