Lookalike audiences look for people similar to your best customers. With changes in privacy and platform automation, their role has evolved: they remain useful as a starting signal, but are no longer the only way to find new customers. The secret lies in the quality of the source list.
The essentials in a definition
Lookalike audiences and audience expansion have changed: the system tends to use more automation and less manual control, and the quality of the source list and the conversion signal remain decisive. It still holds that quality customer lists (with a legal basis) and reliable events, via the Conversions API, improve results. Consult the documentation for current options and the GDPR for list handling.
How they work
The platform analyses characteristics of a source list (customers, high-value buyers, qualified leads) and finds people with similar patterns. The quality of the source list determines the quality of the expansion. Small, outdated or low-quality lists produce weak results.
Garbage in, garbage out.
What has changed
With more automated systems, algorithms explore beyond the defined audience, and lookalikes often work as a suggestion rather than a limit. Privacy restrictions have reduced the precision of some signals. Options and behaviour vary; consult current documentation.
The quality signal still matters.
Lookalike audiences and expansion
- Choose a quality source list
- Update and have consent
- Compare with broad audiences
- Measure customer quality
Good source lists
Highest-value customers and those who bought more than once; leads that became customers; customers with good margins. Avoid low-quality lists, casual visitors or mixtures. Update regularly. Have consent to use them. Test different sources (value, repurchase, category).
Those we want more of are those who already gave the most value.
A lookalike audience is only as good as the customers that originate it.
Test and measure
Compare lookalikes with broad audiences and with interests, in controlled tests. Observe customer quality, not just cost. Test different expansion levels. Combine with good creatives. If performance drops, renew the source list and creatives before abandoning.
Test your reality; the same method does not always win.
Decision table
| What changed | What remains | Action |
|---|---|---|
| More automation in expansion | Signal quality is decisive | Improve events and lists |
| Less manual control | High-value customer lists help | Segment by value |
| Options that change | Need for legal basis | Check consent |
| More aggregated attribution | Incrementality testing | Compare with control |
Common mistakes to avoid
- Outdated lists. They reduce the signal’s usefulness.
- Using lists without a legal basis. It breaches privacy.
- Assuming lookalikes replace good creative. Creative still qualifies the audience.
How to measure the result
Compare cost per new customer, value and retention between approaches.
Limits and exceptions
Options and names change, and automatic expansion reduces transparency. The effect varies by account and volume. Confirm current documentation and avoid conclusions from a single test.
Illustrative example
Illustrative example: a store uses a list of its highest-value customers (with consent) as a signal and compares a campaign with automatic expansion to a more guided one, with the same creative and budget. It measures new customers and value after eight weeks.
In practice: checkpoints
Update customer lists regularly, remove those who asked to leave and segment by value. Compare a campaign with automatic expansion to a more guided one for at least four weeks, with the same creative and budget. Review the quality of new customers, margin and retention, not only initial conversion rate. Document the options you choose, because they may change, and confirm the legal basis of each list before using it. Test different list sources, such as top customers by value versus recent buyers, and watch for audience overlap between campaigns that could raise your costs. Keep the creative variety high, since the system relies on it to find new people.
Frequently asked questions
Do lookalikes still work?
Often yes, as a signal; test against broad audiences.
Which list do I use?
Highest-value customers, updated and with consent.
Do I need large lists?
Good quality and sufficient size; both count.
Are lists still useful?
Yes, especially of high-value, up-to-date customers.
Do I need a legal basis to use lists?
Yes; use only data with an appropriate legal basis and inform data subjects.
Keep high-value, lawful, up-to-date customer lists and reliable events, and compare automatic expansion with more guided approaches in a test.
Sources and recommended reading
- Meta: advertising standards
- Meta for Developers: Conversions API
- General Data Protection Regulation (GDPR)
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.