Lookalike Audiences Start With The Seed List

A lookalike audience is only as good as the list it is built from. Meta takes the people in your source list, finds what they have in common, and goes looking for more people like them. A list full of the wrong customers produces more of the wrong customers.

The setting that matters most is not the lookalike percentage. It is who you put in the seed.

How Meta builds one

You give Meta a source audience: a customer list, website visitors, or people who engaged with your page or videos. The source needs at least 100 people from a single country before Meta will build anything from it.

Then you choose a size between 1 and 10 per cent of the population of the target country. A 1 per cent lookalike sits closest to the source, and a 10 per cent lookalike is broader and looser.

Why the seed goes wrong

Take a retailer that ran a heavy sale last year. A large share of its customer file bought once, at a discount, and never came back. Upload the whole file as a seed and the lookalike learns that the ideal customer is a discount hunter.

Service businesses hit the same problem with leads. A seed built from every form fill includes students, job seekers, suppliers selling to you and people who were never going to buy. The lookalike faithfully finds more of them.

Better seeds

Build the seed from people who did the thing you want more of:

• Customers who bought more than once, or bought at full price.
• Clients whose jobs closed above your average job value.
• Leads your sales team marked as qualified in the CRM, rather than every enquiry.
• For a long sales cycle, the accounts that signed in the last 12 to 24 months.

A smaller, cleaner seed is worth more than a big mixed one. A few hundred real buyers tell the system more about your market than ten thousand newsletter sign-ups.

Value-based lookalikes

If your customer list carries a value for each person, such as lifetime revenue, Meta can build a value-based lookalike that gives more weight to the people who spent the most. For a business with a wide spread of order sizes, this pushes the search toward the top of the customer base rather than the middle.

Lookalikes inside Advantage+

Meta's Advantage+ audience treats the audience you supply as a suggestion, and it can show ads beyond it when it finds better results elsewhere. The lookalike still matters here, because it tells the system where to start looking.

Google retired its similar audiences in 2023. On Google the same job falls to Customer Match lists and the conversion data you send back, and clean seed data does the same work there.

Keep the seed fresh

A customer list exported once and left alone goes stale. People move, change jobs and stop buying, and the business changes what it sells.

Refresh source lists every quarter, and rebuild the lookalike when the offer, the price point or the best customer changes. Keep a short note of what went into each seed, so the next person can tell which list produced which result.

Consent comes first

Every list uploaded to an ad platform has to be one you are allowed to use that way. Check that your privacy policy covers using customer details for advertising. Meta hashes the data before it is matched, but the decision to upload it is still yours.

Across the portfolio, paid runs at a 20x average return on ad spend, and past results do not guarantee future outcomes. Our paid work starts with the inputs, and the seed list is one of the first worth getting right.