Build lookalike audiences from your best customers

A lookalike audience is a group of new people an ad platform picks because they resemble a list you supply, such as your customers. Meta calls them lookalike audiences, and other platforms offer similar options under different names. The common mistake is to seed them with every customer or every website visitor. A better seed is your best customers: those who spent the most, stayed longest or bought again. It suits D2C brands, mobile apps and consumer businesses with enough customer data to build a meaningful seed list.

Why it works

Ad platforms know a lot about the people on their networks. When you give them a seed list, they look for shared patterns and find others who share them. The patterns follow whatever you feed in. A list of one-time discount buyers produces more discount buyers. A list of repeat, full-price customers points the platform toward people with higher lifetime value.

Meta’s own help page makes the same point: results can depend on the quality of the source, and a source built from your best customers may work better than one that includes everyone. Meta also explains that smaller percentage lookalikes match the source more closely, while larger percentages create broader audiences.

How to do it

1. Define what “best” means. Pick one measure that matters to your business, such as total spend in the last 12 months, number of orders, subscription length or in-app purchases.
2. Export a clean list. Pull customers with email, phone and the value measure from your store, CRM or app backend. Remove test accounts, staff and fraudulent orders.
3. Keep the top tier. Sort by your measure and keep the top slice, large enough to meet the platform minimum. Where available, use a value-based lookalike that lets you upload customer value with the list.
4. Check consent and upload safely. Confirm your privacy policy covers ad matching. Use the platform’s upload tool, which hashes the data, or a direct integration with your store.
5. Create more than one size. Build a close match and a broader version. Use ad set location settings to target the right countries.
6. Test against broad targeting. Run the lookalike ad sets alongside a broad audience with the same creative and budget. Compare cost per purchase and new customer rate, and use a holdout where possible.
7. Refresh regularly. Update the seed list monthly or connect it to your store so new best customers are added automatically.

Example

A D2C coffee subscription brand has used a lookalike based on all purchasers for a year. Many of those purchasers bought a single discounted trial bag and never came back. The team exports customers who have kept their subscription for six months or more and uploads that list as a new source. They create a close-match lookalike and a broader one and run both against their existing audience with the same three ads. After several weeks, the new audiences bring in buyers who are more likely to choose a subscription at checkout, so the team moves budget toward them.

A meditation app does the same with users who reached their 30th session, then uses the lookalike for install campaigns.

Metrics to track

• Cost per purchase or install by audience, compared with your broad and existing audiences.
• Share of new customers among buyers. A good lookalike should bring in people who have not bought before.
• Early value per customer. Track 30- or 60-day revenue or retention for each audience, since the goal is customers who stay and spend more.
• Audience overlap. Check overlap between lookalikes and other ad sets to avoid competing with yourself.

When it backfires

• Seeding with the wrong people. A seed full of bargain hunters or low-quality leads teaches the platform to find more of them.
• Stale lists. A seed that never updates misses recent changes in who your best customers are.
• Privacy shortcuts. Uploading data without the right consent or legal basis can breach privacy laws and platform terms.
• Restricted categories. Meta notes that lookalikes are limited or unavailable for some ads about financial products, employment or housing. Check before planning a campaign around them.

Questions people ask

How big should my source audience be?

According to Meta's help pages, a source needs at least 100 people, and Meta generally recommends between 1,000 and 5,000. Other platforms set their own minimums, so check before you build.

Do lookalikes still matter now that broad targeting works so well?

Ad platforms increasingly find buyers on their own with broad targeting and good conversion data. Lookalikes are still a useful signal and a fair test, so compare them with broad targeting in your own account before deciding.

Can I use a customer list without consent?

Uploading customer data to ad platforms is regulated in many regions. Make sure your privacy policy and consent collection cover this use, and follow each platform's rules for hashing and terms.

Sources

• About lookalike audiences (Meta Business Help Center)
• Create a lookalike audience (Meta Business Help Center)