openaso

App Store search ranking factors: what orders the results

Metadata match decides whether an app is a candidate for a query. Behavior — taps and downloads earned on that query — decides where it ranks among the candidates.

What Apple has said

Apple's published guidance on App Store search names two families of signals: text relevance — how the query matches the indexed metadata — and user behavior, meaning taps and downloads earned from search. Everything sharper than that is observation, so this page labels its confidence as it goes.

The observable factors

  • Metadata match and field weight (high confidence): the combined word pool gates candidacy; title words weigh most, then subtitle, then keyword field, with an intact phrase in a visible field a mild extra signal.
  • Download velocity on the query (high confidence): apps that win taps and installs from a search result rise for it; the loop compounds in both directions.
  • Conversion rate from impression to install (high): a result that is shown and skipped is being measured too — icon, screenshots, and rating do ranking work through conversion.
  • Rating mass (medium-high): large healthy rating counts correlate strongly with holding top slots; per-storefront, like everything else here.
  • Engagement and retention (medium, undocumented): rankings shift without metadata changes or download spikes — deletion and retention are widely believed to feed the behavior bucket. Directionally real; unmeasurable from outside.

What does not matter (much)

  • Word order and case in metadata — matching is set-based over the pool.
  • Ratings and behavior earned in other storefronts: each country's index ranks on its own signals.
  • Description text — not indexed for search on the App Store.
  • Bought install bursts without retention: behavior signals decay, and rankings bought that way decay with them.

Watching rankings move

Public rank data covers the top 200 results per query per storefront — the iTunes search API's cap, which is why honest tools report deep positions as not in the top 200 rather than a made-up number. Rankings are per-storefront and move daily; single reads mislead, so measure changes as series.

rank → one keyword, two storefronts
{
  "keyword": "habit tracker",
  "storefronts": [
    { "country": "US", "rank": 12, "totalResults": 200, "prev": 15, "change": "+3" },
    { "country": "GB", "rank": 41, "totalResults": 200, "prev": 38, "change": "-3" }
  ]
}

For attribution, hold one variable still: a metadata release re-indexes within about a day and then settles for weeks as behavior accrues — judge it on the tracked series, not the day-one read.

FAQ

Why does an app with worse metadata outrank mine?

Because metadata only gates candidacy. An app with more downloads, better conversion, and heavier rating mass on that query outranks a better-optimized pool — the fix is winning taps on the queries you already surface for, usually via icon, screenshots, and rating health.

How long after a release do new keywords take effect?

Indexing typically lands within about a day of the version going live. The settling is slower: position among candidates depends on behavior signals that accrue over days to weeks.

Does price affect search ranking?

No direct effect is evidenced. Price moves conversion and downloads, which are ranking inputs — so it acts through the behavior loop, not as a search factor itself.

What does it mean that my app is 'not in the top 200'?

The public search API returns at most 200 results per query, so positions beyond that are unobservable. Tools quoting rank 350 are extrapolating; the honest statement is that the app is outside the measurable window.

Terms

  • text relevance — The metadata-match family of ranking signals: query words against the indexed pool, weighted by field.
  • behavior signals — Taps, installs, conversion, and retention earned on a query — the signals that order apps once metadata makes them candidates.
  • rank — An app's position in one storefront's results for one query, observable to a depth of 200.

markdown mirror: /reference/search-ranking-factors.md · full corpus: /llms-full.txt