How refinement focuses without starving

What refinement controls#

Refinement layers demographic and activity filters over your targeting: age bands, gender and lapsed versus active download behaviour shape the eligible pool before the auction runs. They apply across the exactly 4 App Store placements, so one filter change touches Today tab browsers, Search tab suggesters, searchers and rail shoppers alike. Apple own audience refinement guidance lists the current filter set and their combinations.

Treat filters as seasoning, not the meal: placements, keywords and listings do the heavy lifting, refinement trims the edges. The family view starts at the Apple Ads hub, with placement context in search results.

The reach trade off, quantified#

Every filter multiplies away reach: two age bands plus one gender can halve eligible impressions before bidding even matters, and thin pools then deliver lumpily with noisy reads. Small markets feel this first, since their pools start shallow. Rule of thumb: keep every filtered ad group large enough to earn steady daily impressions, or broaden before judging.

Market depth planning lives in our 91 markets guide, with location pooling in location targeting. Confirm filter availability per market in the developer promotion docs, since options differ by storefront.

Start broad, refine on evidence#

Launch unfiltered or lightly filtered, collect a fair sample of taps and installs, then cut only the segments the data condemns: the age band that never converts, the activity slice that bounces. Evidence led refinement compounds, while assumption led filtering bakes prejudice into structure. Revisit quarterly at minimum, because audiences drift constantly with seasons, launches and school calendars.

Prove segment reads with splits from our placement reporting guide before locking them into structure. Harvesting rhythm in keyword harvesting pairs well with refinement reviews.

Age and gender, handled carefully#

Use age bands to protect relevance, not to stereotype: exclude only where the product genuinely cannot serve, such as age gated categories, and prefer bid or budget weighting over hard exclusion elsewhere. Gender filtering suits only apps with genuinely gendered demand; for most apps it halves learning to prove a hunch. Document every exclusion reason in the media plan so future reviewers can challenge it.

Where demand truly splits, mirror structure instead: separate campaigns per segment with tailored listings via custom product pages, and read them in customer type splits.

Activity signals that matter#

Download activity filters separate fresh hunters from lapsed reinstall candidates, and the two need different promises: hunters need education, returners need reasons to come back. Route returners to pages that acknowledge the update, the new season and the fixed flaw head on. Generic pages waste activity signals entirely, so match every activity slice to its own tailored page.

Winback page craft draws on our reengagement guide, with account splits in account structure keeping hunter and returner budgets apart for clean reads.

Review the split monthly: hunter volume funds growth while returner efficiency funds profit, and the right balance shifts with seasons and launches. Rebalance deliberately, never by drift. Seasonal splits guide the calendar.

Report hunter and returner economics separately to stakeholders: blended CPA hides whether growth or efficiency moved. Separate stories earn smarter budgets. Split reporting keeps both honest.

Mistakes that strangle delivery#

Stacking every filter at launch, copying web audience habits into store buying, and filtering small markets like large ones: the classic triple that leaves campaigns becalmed. Each feels precise and each starves the auction of the volume it needs to learn. Precision without volume is just silence with better labels, and labels never paid a developer.

Fix with a filter budget: at most two active filters per ad group at launch, each removable within a week on data. Log every filter change with its reason and review date so removals happen on schedule rather than never. Persistent delivery puzzles go to our team for a structure audit.

Frequently asked questions#