A lookalike audience in Meta ads is a new audience built from a source list you already trust, usually customers or qualified leads. You create it in Audience Manager from a custom audience, choose a percentage between 1% and 10%, and add it to an ad set, which supplies the location. Meta then finds people who share characteristics with the source.
Steps to create one
- Build or pick the source: a custom audience of customers, or of leads your CRM marked as qualified.
- In Audience Manager, create a lookalike audience and select that custom audience as the source.
- Choose the percentage. One percent is the closest match, 10% the widest.
- Save it. Meta takes up to 3 days to populate the audience, and ads can run with it in the meantime.
- Add it to the ad set. The estimated size appears once the lookalike is used there, because the location now comes from the ad set and not from the moment of creation.
You can save up to 500 lookalikes from a single source, which is more than any lead gen account needs. Adding several lookalikes to one ad set makes Meta target people in any of them, which is a way to widen reach without stacking ad sets.
The settings that decide the match
| Setting | Rule as Meta documents it |
|---|---|
| Source size | at least 100 people, with 1,000 to 5,000 recommended |
| Lookalikes per source | up to 500 |
| Audience size | 1% to 10% of the population of the ad set’s locations |
| Location | taken from the ad set, not chosen at creation |
| Populating | up to 3 days, and ads can run meanwhile |
| People in the source | excluded from the lookalike automatically |
| A saved audience as source | not allowed |
Two of those rows change how you plan. The source is always a custom audience, never a saved one, so a targeting combination cannot be the seed. And because the source people are excluded, a lookalike built from your customer list will not retarget those customers, which is usually what you want.
The source is the whole decision
A lookalike teaches Meta what your source looks like. Point it at every form fill and it learns what a form filler looks like, which is the wrong person to find more of. Point it at the leads that qualified or closed and it learns something closer to your buyer.
Meta’s own advice is that a source of your best customers can beat one that includes all of them. In a lead gen account that usually means a list of a few hundred qualified leads rather than a few thousand raw ones. The minimum of 100 is low enough that a small, clean list still works, and Meta’s Leads Center can build custom and lookalike audiences from your leads once you have passed 20 leads on Meta.
Which leads qualified is decided in your CRM, not in Ads Manager. If the CRM cannot tell you, the lookalike is guessing along with you.
Lookalikes and Advantage+ audience
Meta switches Advantage+ lookalike on automatically for new, duplicated and draft ad sets that use lookalikes built from app, website or customer list custom audiences. In Leads campaigns, lookalikes sit under custom audience inclusions in the suggestions box, which makes them a suggestion rather than a limit.
That distinction decides your test. A 1% lookalike added as a suggestion is a hint: delivery can go outside it when Meta expects better results. To make the lookalike itself the constraint, choose Further limit the reach of your ads, select the audience again and uncheck Use as a suggestion. The account structure view of where audiences live explains why this sits at ad set level and not in the campaign.
Where lookalikes are limited
Special Ad Categories come first. Meta lists lookalike audiences as limited or unavailable for housing, employment and financial products and services campaigns in the countries it names, and Advantage+ lookalike as unavailable there.
Sources can also be flagged. If the custom audience behind a lookalike is flagged for information Meta’s terms do not allow, the lookalike inherits the flag, and the fix starts at the source rather than at the lookalike.
How to judge a lookalike
Do not judge it on cost per lead. A lookalike built from qualified leads can produce a higher cost per lead than broad delivery and still be the better audience, because the leads are closer to the ones that pay. Judge it on qualified leads and on cost per qualified lead, and read those per creative and landing page so the audience is not credited for a creative that would have worked anywhere.
Run it as a control in at least one arm of the test, with the lookalike as a hard inclusion, so the comparison is between the lookalike and something, not between the lookalike and nothing. The rest of the setup is in campaign setup, budget and delivery, and what a lookalike is at the glossary level is in lookalike audience.