How interest clusters find buyers

What interests actually are#

Activity clusters, not identity tags: interests reflect what people repeatedly do, which predicts better than what they once liked. Treat each interest as a behavioural hypothesis to test, never as a persona to believe.

Detailed targeting consolidated in mid-2025, so current options differ from old guides: verify live inventory in Ads Manager before planning. Meta own targeting options guide lists what exists today. Start at the Meta Ads hub.

Prefer mid-size interests with clear commercial intent over mega-interests that mean everything. Broad targeting covers the alternative.

Mine interest ideas from customer language: reviews, support tickets and sales calls name the activities buyers actually do. Mirror their words back as stacks. Customer language seeds better stacks.

Sizing the stack#

Stack related interests to viable size: too narrow starves learning, too broad drowns the hypothesis. Estimate reach per stack before launch and fund each ad set to roughly 50 events a week.

One theme per ad set keeps reporting legible: fitness buyers separate from fashion browsers, always. Mixed stacks produce blended mush. Consolidation sets the ad-set count.

Record stack definitions with dates: interest options change and old names mislead future audits. Ask our team for a stack audit when performance drifts.

Testing interests honestly#

Test one stack against broad with identical creative and budget, and read incremental CPA over two weeks. Interest tests without a broad control prove nothing except that spending spends.

Promote winning stacks into dedicated budgets and retire losers quarterly. Interest winners decay as audiences tire and options shift. Clean tests set the read rules.

Log every test with hypothesis, stack, spend and verdict: the library compounds. Test logs compound like creative libraries.

Interest versus broad#

Interests win where signal is thin and guidance helps: new pixels, niche products, tight budgets that cannot fund roaming. Broad wins where purchase signal runs deep enough to navigate alone.

Graduate stacks into broad as events accumulate: today's manual winner becomes tomorrow's suggestion seed. Advantage+ audience absorbs proven stacks.

Never run identical stacks against broad simultaneously without exclusions: overlap starves both. Overlap guards keep tests clean.

Schedule the graduation review with budget planning: stacks that earned scale deserve funding windows. Promote winners formally, not by accident. Graduation belongs on the calendar.

Retiring dead interests#

Kill interests that spend without learning past 3 to 5 times target CPA with zero conversions. Sentimentality about clever stacks is just expensive nostalgia.

Review stacks quarterly against current options: consolidated or removed interests silently change what ad sets buy. Fatigue curves distinguish tired creative from dead targeting.

Recycle budgets from kills into broad tests, not into lookalike stacks of the same dead idea. Scaling rules reallocate the winnings.

Hold killed stacks in a graveyard sheet with the reason attached: dead ideas resurrect seasonally and deserve a fair rehearing. Graveyards prevent repeat mistakes. Rested stacks sometimes revive.

Interests in the full mix#

Interests are one input among many: creative, offer, signal and budget usually move CPA more than the next stack tweak. Fix the big levers before micro-optimising clusters. Meta diversification research shows why creative usually moves more.

Pair interest prospecting with engagement retargeting so warm traffic compounds cold tests. Prospecting splits structure the mix.

Report interest performance inside the account story, never as isolated wins: stacks serve strategy, not the reverse. First-campaign setup places interests correctly.

Cap interest testing at a fixed share of prospecting: endless stack experiments starve proven broad delivery. Ten to twenty per cent tests, the rest earns. Testing budgets protect winners.

Frequently asked questions#