ASO Patents

Patent US 2013/0325892 A1 — “Application Search-Query Classifier” (Apple Inc.)

Key facts	Details
Publication	5 Dec 2013
Inventors	Catherine A. Edwards et al.
Core IPC	G06F 17/38 → information-retrieval ranking

1. What problem does it solve?

With millions of apps, a plain keyword search often drowns the user in irrelevant results. The patent teaches a server-side classifier that first infers the intent behind a query and then runs a category-specific ranking strategy.

2. Three query categories

Category	Apple’s definition (verbatim)	Typical signals
Navigational	“in which it is estimated that a user is searching for a particular app by name”	Capital letters, quotes, few words, very skewed historical download distribution
Functional	“in which it is estimated that a user is searching for an app that performs a function”	Lower-case, longer phrases, broad tail of downloads
Browse	“in which it is estimated that a user wishes to explore a variety of apps.”	Modifiers like “top”, “best”, “free”, large result set

3. How the classifier works

1. Term analysis – number of words, capitalisation, presence of “pro/lite/best” tokens.
2. Pre-search probe – count exact-title hits and total hits.
3. Empirical distribution – look at past sessions: for the same query, how many distinct apps were downloaded and how concentrated were those downloads?
   Example logic: if ≥ 50 % of downloads go to one app and ≤ N distinct apps exceed a cumulative 90 % share, flag as navigational (see flowchart Fig. 7).
4. Assign category → pick ranking recipe:
   • navigational ⇒ exact-title match + string edit distance;
   • functional ⇒ synonym/metadata match + popularity weight;
   • browse ⇒ popularity & rating first, text match later.

4. Re-ranking & feedback loop

After results are shown, the system logs view → click → download chains, feeds them back into the empirical database, and periodically recomputes the thresholds so the classifier keeps learning.

---

Connection to ASO (App Store Optimization)

Insight from the patent	Practical ASO action
Exact-match dominance in navigational searches. If a single app hoards most downloads for a query, Apple treats that query as “find that app fast”.	Title hygiene: keep your brand name intact in the first 30 chars; avoid unnecessary prefix/suffix variants.
Functional queries rely on semantic metadata & popularity.	Keyword field & description: target feature words (“invoice scanner”, “guided HIIT”) and support them with healthy ratings and download velocity to win the functional ranker.
Browse queries boost apps with high engagement metrics.	Rating velocity & installs: campaigns that spike reviews or installs around “top”, “best”, “free” keywords lift browse positioning.
Classifier learns from post-result behaviour. Higher click-through and install rates on your listing feed back as positive signals that can upgrade its category scores.	Creative optimisation: screenshots & preview video that lift tap-through will indirectly raise rank even for the same keywords.
Query tokens like “pro”, “lite”, version numbers are navigational hints.	Consistent SKU naming: align your paid, lite and seasonal SKUs so users (and Apple) can map all to the same brand query rather than splitting relevance.

Bottom line: the classifier is Apple’s early blueprint for the modern App Store search stack. Understanding which bucket your target keywords fall into lets you tailor metadata, creative assets and growth tactics to the specific ranking algorithm Apple will apply.