How to Use Lookalike Audiences on Meta Ads

By Kristians, Founder, AutoAdy. Six years buying Meta ads — first for agency clients, then for his own ecommerce and lead-gen brands.

Last updated:

What changed: Turned the size comparison into a table, made the H2s question-form, and added the overlap calculator.

Lookalike audiences find people resembling your customers. 1% lookalike targets top 1% most similar (~2.1M in US) with lowest CPA. Source quality beats size — 500 high-value customers outperform 5,000 random leads.

Key Takeaways

  • ✓1% = lowest CPA (2.1M in US); 5% = 20-30% higher CPA; 10% = near-broad
  • ✓500 high-LTV customers outperform 5,000 random leads as seed audiences
  • ✓Stack 1%, 1-3%, 3-5% in separate ad sets for controlled scaling

Which lookalike audience size performs best?

Start at 1% and go broader only once it saturates — frequency above 3.0 is the signal. By 10% a lookalike behaves close to broad targeting, so the extra targeting work stops paying for itself.

Lookalike tiers, US audience size and relative CPA against the 1% tier.
TierUS audience sizeCPA versus 1%
1%~2.1MBaseline — best CPA
5%~10.5M20-30% higher
10%~21M40-60% higher, close to broad

What makes a good lookalike seed audience?

Best: highest-LTV buyers > all buyers > high-intent leads > all leads > visitors. 500 top buyers beat 10K email lists. Min 100 people, 1K+ for stability.

Stacking Technique

Three ad sets: 1%, 1-3% (excludes 1%), 3-5% (excludes 1-3%). Budget: 50/30/20 split. Shift budget as tiers saturate. Scales without diluting best audience.

What we see in AutoAdy accounts

Figures below come from Meta ad accounts AutoAdy has audited or monitors. They are directional, not a controlled study.

  • •1% lookalikes deliver 35% lower CPA than 5% across AutoAdy accounts
  • •Top 20% LTV sources produce 42% better results than all-customer sources
  • •AutoAdy detects saturation and recommends next-tier expansion

Sources

FAQ

Common questions

What size lookalike?

1% for best CPA. Expand to 3-5% when 1% frequency exceeds 3.0. 10% barely outperforms broad targeting.

Minimum source size?

100 in one country, 1K+ for stability. Quality over quantity — 500 high-value buyers beat 5K random leads.

Still effective in 2026?

Yes, less dominant post-iOS 14.5 but still 15-25% better than interest targeting. Combine with Advantage+ for best results.

Related

Scale Meta Ads Without Killing ROAS

Scale Meta ad spend without tanking ROAS.

How to A/B Test Facebook Ads

A/B test Facebook ads the right way. Minimum sample sizes, one-variable testing, and the 95% confidence threshold that separates winners from noise.

Free Audience Overlap Calculator

Check that stacked lookalike tiers are not competing.

Ready to try lookalike performance tracking that actually works?

Start free. Connect your Meta account. See your wasted spend in 60 seconds.

Start freeSee how it works