Campaign setup, budget and delivery

Custom audience vs lookalike audience: which one should a lead generation campaign use?

A custom audience is the list you already own, a lookalike is new people Meta finds from it. Lead generation needs both, and the seed decides which one works.

Published

A lead generation account usually needs both, because they do different jobs. A custom audience is a list of people you already know, and a lookalike audience is new people Meta finds by comparing them with that list. Prospecting runs on the lookalike, retargeting and exclusions run on the custom audience, and the list you seed decides how well the lookalike performs.

Two audiences, two jobs

A custom audience holds people you have a relationship with, built from data you upload or that Meta already has: a customer file, your own leads, visitors to your site, people who engaged with the Page or with an ad. Everyone in it is already known to you.

A lookalike is built the other way round. You name a source audience, Meta studies signals such as demographics, interests and behaviors inside it, and it offers you a new group of people who resemble that source. The percentage you choose sets how close the resemblance has to be.

Custom audienceLookalike audience
Who sits in itpeople you already knowstrangers Meta selected as similar to a source
What builds ityour own data or your existing engagementa custom audience you name
How large it can bewhatever your data supports1% to 10% of the population in the ad set’s locations
Usual job in lead genretargeting, exclusions, the seedprospecting
Can it seed a lookalikeyesno

Because every lookalike begins as a custom audience, the choice is rarely one against the other. The decision that changes results is who goes into the source.

Meta’s limits before you build anything

The rules Meta publishes decide most of the structure for you.

SettingWhat Meta requires
Source listno fewer than 100 people, with 1,000 to 5,000 the range it suggests
Lookalikes from one sourceas many as 500
Audience sizebetween 1% and 10% of the people in the ad set’s locations, with 1% the closest match
Locationset by the ad set now, so it is no longer a choice at creation
Fill timeup to three days, and ads may run while it fills
The people in the sourceremoved from the lookalike automatically
A saved audience as sourcerefused

Two consequences follow. A new lookalike has no estimated size in Audience Manager until you put it in an ad set, because the location has moved to the ad set. And several lookalikes inside one ad set mean Meta can reach anyone in any of them, which is a cheap way to widen reach without building a new ad set per percentage.

Seed with the leads that qualified

What Meta looks for is defined by the source. Its own advice is that a list of your best customers can beat a list of all of them, and in lead generation the equivalent is leads that qualified or closed rather than every form fill.

Say you hold 2,000 leads and 300 qualified. A lookalike of the 2,000 teaches Meta the shape of someone who fills in a form. A lookalike of the 300 teaches it the shape of someone worth calling, and 300 is comfortably above the 100-person floor. Once you have produced more than 20 leads on Meta, Leads Center can also build custom and lookalike audiences directly from them.

Qualification happens in your CRM, not in Ads Manager, so the seed list is a reporting task before it is an audience task. Keep the ad ID on every record and the list can be rebuilt whenever you need a fresh source.

What only a custom audience can do

Three jobs have no lookalike equivalent.

Retargeting: people who clicked, opened the form or visited the landing page without submitting. You already know them, so there is nothing for Meta to find.

Exclusions: your current customers, or the leads the sales team is already calling, should not see the same offer again.

Building the seed: a clean list of qualified leads is the input that makes any lookalike worth running.

For how many ad sets a weekly budget can carry, see Facebook ads account structure for lead generation and the other campaign setup questions.

Advantage+ lookalike turns it into a suggestion

Meta enables Advantage+ lookalike automatically on new, duplicated and draft ad sets whose lookalikes come from app, website or customer list custom audiences. With it on, Meta may reach people outside the lookalike when it expects better performance.

In Sales, App promotion and Leads campaigns, a lookalike added through Add suggestions lands under custom audience inclusions, which makes it advisory. To turn it into a boundary, use the option to further limit the reach of your ads, pick the lookalike again, and clear the setting that treats it as a suggestion. Note also that a lookalike cannot be excluded while Advantage+ audience is switched on.

So a 1% lookalike added as a suggestion is guidance, not a fence. If you want to know whether the lookalike itself earns its place, run it as a control in at least one arm of a test.

Categories and flagged sources

Housing, employment and financial products and services campaigns face the first restriction: for US-based advertisers, or campaigns reaching the US, Canada or certain European countries, Meta lists lookalike audiences as limited or unavailable, and Advantage+ lookalike as unavailable. A mortgage or insurance offer that falls under housing or financial products follows those rules even where the same offer has no category in another market.

The second restriction runs through the source. A custom audience flagged for information Meta’s terms do not permit takes its lookalike down with it, and the repair starts at the source, not at the lookalike.

Read the result on your own cost per lead

A lookalike built from leads that never qualified finds more people like them. It can look efficient inside Ads Manager while the sales team keeps rejecting what arrives, so the number that settles the question is cost per lead on leads you actually received, read per ad and per landing page. CPL vs CPA for lead generation covers which of the two belongs to that decision.

A workable order for the test:

  1. Run the lookalike beside a broad control with the same creatives and the same landing pages.
  2. Wait until both sides have spent enough for a CPL to mean anything.
  3. Compare the leads you received, not the platform’s count.
  4. Check cost per qualified lead before scaling the winner, since a cheap lookalike of unqualified leads is an expensive audience.

The budget mechanics that keep both arms fed, from bid strategy to how many ad sets a weekly budget supports, sit in CBO vs ABO and cost cap vs bid cap. The definition of the audience, with the same limits repeated, is in the lookalike audience entry.

Questions on this topic

All questions on this topic →

FAQ

What is the minimum size of a lookalike source?+

Meta requires at least 100 people in the source audience and recommends 1,000 to 5,000. In lead generation a smaller source built from qualified leads often beats a larger one built from every form fill, because Meta looks for more people like the source.

Can I use a saved audience as a lookalike source?+

No. Meta does not allow a saved audience as the source. Use a custom audience built from a customer list, your leads, website traffic or engagement.

How long does a lookalike take to populate?+

Up to 3 days according to Meta, and ads can run while it fills. New lookalikes show no estimated size in Audience Manager until you use them in an ad set, because the location now comes from the ad set rather than from the audience.

Can I exclude a lookalike from an ad set?+

Not while Advantage+ audience is on. Meta says you cannot exclude a lookalike in that setup, so if exclusions matter, build the ad set without Advantage+ audience.

Do special ad categories affect lookalikes?+

Yes. For housing, employment and financial products and services campaigns from US advertisers or reaching the US, Canada or certain European countries, Meta lists lookalike audiences as limited or unavailable, and Advantage+ lookalike as unavailable.

Ready to find out what
converts, for real?

Free to start.

Start free →