What is Lookalike Audience?

How lookalike audiences actually work

You supply a source: a customer list, a pixel-based audience, or people who took one specific action. The platform then analyses patterns across its own user data and assembles a matching audience using signals it will never disclose to you.

Meta lets you set size from one to ten percent of a country’s users, where one percent is the tightest match on the smallest pool and ten percent is broad enough to sit closer to interest targeting than genuine similarity. Google works from customer match lists similarly. Both need roughly a thousand matched records to function.

Where teams get lookalikes wrong

Seeding with the wrong list. All purchasers includes your one-time discount hunters, so a lookalike built on them obediently finds more discount hunters, which is why you should seed on high-value repeat customers even when that list is considerably smaller.

Second, going too broad too early, because a ten percent lookalike on a young account is barely distinguishable from open targeting. Start at one to two percent.

Third, letting the seed go stale. A list uploaded eighteen months ago describes a customer base you no longer have, so refresh quarterly and build separate lookalikes for genuinely distinct segments rather than averaging two customer types into one blurred audience.

What good looks like in India

Build the seed from customers who bought twice or more at full price, then hash and upload phone numbers rather than only emails, because match rates on Indian phone numbers typically run much better than on email addresses and formatting inconsistency will otherwise cost you a chunk of the list.

Then layer geography on top. A one percent lookalike across all of India skews toward wherever population density is highest, which may be nowhere you ship profitably. Restricting to serviceable cities usually beats any change to the seed.