Meta audiences and targeting

Pre-2021, Meta targeting was surgical: obscure interests, narrow demographics, custom combinations. Post-iOS 14, signal loss has made broad targeting often win. The platform's AI now does most of the targeting - your creative does the rest.

Audience types

Broad

Country + age range. Let Meta's AI find the conversions. Post-iOS 14 best practice for most advertisers.

Interest targeting

Target people interested in specific pages, topics. Less precise than it once was. Still useful for niche products.

Lookalike audiences (LAL)

Meta finds people similar to your seed audience (customers, email list). 1-3% LALs are tightest; 4-10% broader. Still work but less precise than they used to.

Custom audiences

Your own data uploaded. Website visitors, customer list, video viewers, engaged IG/FB users.

Saved audiences

Combinations of interests + demographics you save for reuse.

The broad vs narrow debate

• Broad + great creative = Meta AI finds buyers. Usually the winner.
• Narrow interest = you guess who buyers are. Usually worse.
• Exception: genuinely niche products (hobby-specific, B2B roles) where broad would waste impressions.

Exclusions

• Exclude current customers from prospecting
• Exclude recent purchasers (30-90 days) from retargeting
• Exclude email subscribers from newsletter-signup campaigns

Geography

Default to the country you ship to. Narrow further (state, city, radius) only if your offer is local or you've proven a geographic pattern.

Demographics

Age range, gender. Only narrow if you have real data that a demographic doesn't convert. Most advertisers over-narrow here and miss buyers.

Placements

Let Meta choose automatic placements 90% of the time. Manual placement selection usually limits the algorithm for no good reason.

What to do with this

• Default to broad targeting + strong creative, the creative does the audience segmentation now that iOS 14 killed most precise targeting
• Use lookalikes based on purchasers (not leads) wherever possible, purchaser lookalikes convert 2-3x better than lead lookalikes
• Don't narrow by age and gender without real data supporting it, most advertisers over-narrow and miss 40% of their buyers
• Let Meta use automatic placements by default, manual placement selection limits the algorithm without adding signal
• Re-test your "audience assumptions" every quarter, last year's winning audience segment is often this year's decliner