A release owner sees twelve recent complaints containing "crash" and asks whether the new build is worse. The number can justify reading those reports. It cannot, by itself, measure crashes per user, prove a regression or describe every review of that version.

The useful first step is to define a bounded review comparison that another person can reproduce.

Disclosure: ApiVault Labs develops the App Store Reviews Scraper discussed here. This guide was prepared with AI assistance and checked against production input schema 0.3.1 on November 17, 2026. No live collection was started for this article.

Write the comparison contract first

Keep app ID, selected storefronts, source sort, collection time, date window and delivery limits with the export. The Actor's updatedAt comes from Apple's source update timestamp; it is not verified first-publication time. A review updated after release may describe an earlier experience.

Start with one app and one storefront. This example uses a public app ID, not a synthetic review dataset, and does not promise that a particular complaint exists today:

{ "appInputs": ["310633997"], "countries": ["us"], "sort": "newest", "dateFrom": "2026-11-01", "dateTo": "2026-11-17", "maxReviews": 100, "maxResults": 100, "maxPages": 10, "includeAuthors": false } 

The date-only end includes the whole UTC day. A daytime collection cannot contain updates that have not happened yet. Save the actual retrieval time separately.

After collecting, inspect which version values the source returned. Compare exact observed version strings; do not assume the current app version labels every available review. Missing version stays its own unknown group.

Preserve the denominator you actually collected

Apple's public RSS window exposes at most ten pages of fifty reviews per app, storefront and selected sort, and may expose fewer. Raising maxResults does not expand that source window. Selective filters can return zero even when the app has many historical reviews.

For an exploratory version comparison, collect an unfiltered sample and retain all stars. If you later calculate the share of low-star reviews, name it "low-star share in this collected sample." A run using ratings: [1, 2] supplies a reading queue, not the denominator for all collected ratings. Neither statistic is a crash rate or an estimate of affected users.

Keyword matches are case-insensitive substrings over title and text. "No crash now" can match "crash." Read the original context before assigning a theme or opening a defect.

Keep app identity with review identity

Deduplicate using (appId, reviewId). The same review can appear through multiple storefronts; the Actor retains the first-delivered country. That country identifies a storefront, not reviewer nationality or residence. A combined multi-country run therefore does not provide independent country totals.

Keep the same review on different app IDs separate. When joining old and new exports, an identical ID with changed text, stars or timestamp needs an explicit rule. Compare collections separately or select a documented observation; silently merging them makes the denominator unstable.

Inspect collection health before briefing the team

Read SUMMARY, COVERAGE and ERRORS from the run key-value store. A successful process can still have partial collection. FEED_WINDOW_SCANNED describes the configured feed window, not complete historical coverage. Save gaps, caps and partial statuses alongside the dataset.

Give a human release owner a few source-linked examples, exact version groups and a next check in owned crash telemetry or support records. Developer replies are not provided by this RSS implementation. Review links lead to the app review list, not a guaranteed single-review permalink.

Current published review event fee: $0.02 per 1,000 useful delivered rows, plus user-paid Apify resources; no startup event. One hundred rows means $0.002 in review fees before resources. Empty or filtered runs can still consume resources. This article does not measure views, conversions or product demand.

App Store Reviews Scraper and current limits