How Advantage+ audience finds buyers

Suggestions steer, controls bind#

Suggestions are hints, not fences: interests, age bands and lookalike seeds tell delivery where to start looking, and the system roams beyond them when signals point elsewhere. Treat every suggestion as a starting rumour, never as a boundary.

Controls are fences: location, minimum age, language and custom exclusions genuinely restrict who can see ads. Put regulatory and commercial non-negotiables in controls, hunches in suggestions. Meta own audience controls guide lists what binds. Start at the Meta Ads hub.

Audit the split quarterly: suggestions that never get explored are decoration, while controls that never bind are risk. Controls versus suggestions details the full mechanic.

Map suggestions to funnel stages: broad hints for prospecting, proven-buyer seeds for closing. Stage-matched suggestions waste less roaming. Stage-matched hints convert.

The default-on reality#

New campaigns default to Advantage+ audience, so opting out is deliberate: switching to original audiences needs a reason written down, not a habit carried over. Defaults compound across accounts into strategy by accident.

Haus tested 640 advertisers and found Advantage+ edging platform ROAS while most brands won incremental ROAS on manual. Neither side always wins, which is why the default deserves a challenger. Haus breakdown reads the evidence.

Keep one manual prospecting cell beside broad delivery until broad beats it twice on incremental ROAS. Ask our team to set the challenger protocol.

What good suggestions look like#

Strong suggestions describe proven buyers: customer lists, qualified-lead seeds, engaged viewers, past purchasers. Weak suggestions describe demographics you hope will buy. The system amplifies evidence, not aspiration.

Refresh seeds as the business changes: last year's buyers mislead this year's model. Seed hygiene is audience hygiene. Custom audiences build the seeds.

Layer one suggestion type at a time when diagnosing: stacked suggestions blur the read on what worked. Clean tests isolate the signal.

When broad fails#

Broad fails where eligibility is narrow: regulated categories, tight geographies, age-gated products. It also fails where signal is thin, because roaming without evidence is guessing with budget.

Constrain with controls first, then manual audiences, and only then blame the format. Most broad failures are signal failures wearing an audience costume. Exclusions tighten the fences.

Document every broad failure with the control set attached: patterns across failures teach faster than any single win. B2B playbooks live in constrained mode permanently.

Migration without drama#

Move ad sets to Advantage+ audience one at a time, holding budget and creative constant, and read incremental ROAS for two weeks before migrating the next. Wholesale flips corrupt every read at once.

Keep the manual originals paused, not deleted, until broad wins twice: rollbacks beat rebuilds. Name migrated sets clearly with dates. Consolidation governs the end state.

Re-migrate quarterly as signal matures: audiences that failed broad at 100 events may fly at 1,000. Scaling rules set the review rhythm.

Measuring the switch#

Judge on incremental ROAS and new-customer rate, never platform ROAS alone: broad delivery finds existing customers efficiently, which flatters blended numbers while acquisition quietly worsens.

Split new versus existing in every report, and hold out periodically to prove the mix. Lift testing settles the argument properly.

Report learning status beside results: Learning Limited broad delivery is a budget verdict, not an audience one. Meta learning phase guide defines the thresholds. Learning maths sets the minimums.

Calendar migration reviews with budget planning: audience moves need funding windows, not Friday whims. Planned moves beat panicked ones. Planned migrations stick.

Frequently asked questions#