Google Play Reviews & App Feedback Analyzer

Collect bounded public Google Play review records for one or more Android apps. The current release uses the public Store HTML page, preserves locale and market parameters, emits source diagnostics, and attaches optional shared-core analysis without coupling analysis to collection.

What it collects

• Review ID, star rating, localized date text, review text, and helpful count.
• Optional developer reply text and date when present in the public card.
• Language and country used for the request.
• A machine-readable diagnostic record for every app, including HTTP status, response size, and parsed-card count.
• A normalizedFeedback object on each review, validated by the shared source-neutral feedback contract.
• Shared analysis, actionable-feedback clusters, per-app aggregate reports, and optional observational release comparisons.
• One APP_REPORT_<app-id> report in key-value storage per processed app when aggregation is enabled.

The public page currently exposes a bounded server-rendered sample, not complete review history. Browser expansion remains deferred; shared deterministic analysis and aggregation are enabled by default.

Who it is for

Product managers, mobile QA teams, support teams, and app developers who need a bounded Google Play review sample converted into traceable product-feedback signals.

Key capabilities

• Collect public Google Play reviews for multiple Android package IDs with language and country provenance.
• Preserve source facts while adding validated normalized feedback and deterministic shared-core analysis.
• Surface recurring actionable feedback as stable issue/feature clusters linked back to review IDs.
• Produce one aggregate report per app and an optional before/after release-impact report.

Raw reviews versus analysis

Review text, ratings, dates, helpful counts, replies, and source diagnostics are collected facts. normalizedFeedback, analysis, clusters, rankings, severity, and release-impact signals are analytical interpretations. Analytical issue labels are user-reported signals, not confirmed product defects.

Input

Field	Required	Default	Description
mode	no	reviews	reviews for normal collection or releaseImpact for a bounded before/after report.
appIds	yes	["com.todoist"]	Android package IDs, up to 20.
language	no	en	Two- or three-letter Google Play language code.
country	no	US	Two-letter Google Play market code.
languages	no	[language]	Language list used by releaseImpact; one public Store request per combination.
countries	no	[country]	Country list used by releaseImpact; one public Store request per combination.
maxReviewsPerApp	no	50	Hard cap from the parsed server-rendered sample, 1–500.
sort	no	mostRelevant	mostRelevant or newest; recorded for diagnostics in the current HTML path.
useBrowserFallback	no	false	Reserved for the later browser-expansion phase.
requestTimeoutSecs	no	30	Per-request timeout, 5–120 seconds.
debug	no	false	Emit normalized-input details through debug logging for local troubleshooting.
analysis	no	enabled	Shared analysis settings: enabled, outputLanguage, and maxAttempts.
aggregation	no	enabled	Cluster and report settings, including optional observational release comparison.
release	releaseImpact	—	Release version and ISO releasedAt timestamp.
daysBefore	releaseImpact	14	Calendar days before the release boundary.
daysAfter	releaseImpact	14	Calendar days after the release boundary.
maxReviewsPerPeriod	releaseImpact	100	Request cap for each language/country slice.

Example:

{
    "mode": "reviews",
    "appIds": ["com.todoist", "com.zing.zalo"],
    "language": "vi",
    "country": "VN",
    "maxReviewsPerApp": 25,
    "sort": "mostRelevant",
    "useBrowserFallback": false
}

Release-impact example:

{
    "mode": "releaseImpact",
    "appIds": ["com.example.app"],
    "languages": ["en", "vi"],
    "countries": ["US", "VN"],
    "release": { "version": "4.2.0", "releasedAt": "2026-07-20T00:00:00.000Z" },
    "daysBefore": 14,
    "daysAfter": 14,
    "maxReviewsPerPeriod": 100
}

Output fields

Field	Description
recordType	review, sourceDiagnostic, feedbackCluster, productFeedbackReport, or feedbackImpactReport.
appId	Android package ID.
reviewId	Public Google Play review ID, for review records.
rating	Integer star rating from 1 to 5.
reviewDateText	Locale-preserved date text from the public page.
text	Public review text, for review records.
helpfulCount	Helpful count when exposed, otherwise null.
developerReply	Optional reply object with presence, date text, and text.
source.language	Language request parameter.
source.country	Country request parameter.
diagnostics.httpStatus	HTTP status for the Store response.
diagnostics.responseBytes	Response size in bytes.
diagnostics.collectedAt	Collection timestamp for the source response.
diagnostics.parsedReviewCount	Number of unique review cards parsed before the cap.
diagnostics.collectionMode	Current value is html.
error.code	Machine-readable source error code when collection fails.
normalizedFeedback	Source-neutral feedback object validated by shared core.
analysis	Shared-core analysis result, when analysis is enabled.
product.productId	App ID on aggregate and impact records.
product.productType	Neutral product type; app for Google Play aggregate records.
clusterId	Stable cluster ID on feedback-cluster records.
canonicalIssue	Canonical issue or request title for a cluster.
feedbackType	Cluster feedback type, such as bugReport.
topics	Topics attached to a cluster or release-impact comparison.
mentionCount	Number of reviews in a cluster.
uniqueReviewCount	Number of distinct source review IDs in a cluster.
languages	Languages represented in a cluster.
countries	Countries represented in a cluster.
affectedVersions	App versions represented in a cluster.
firstSeenAt	Earliest source date represented in a cluster.
latestSeenAt	Latest source date represented in a cluster.
severity	Analytical severity estimate.
clusterConfidence	Bounded cluster confidence estimate.
statistics	Counts, rating, language, country, and version summaries.
statistics.reviewsCollected	Collection count in an app report.
statistics.reviewsAnalyzed	Successful-analysis count in an app report.
statistics.actionableReviews	Actionable-review count in an app report.
statistics.averageRating	Average source rating in an app report.
statistics.beforeReviews	Before-window review count.
statistics.afterReviews	After-window review count.
statistics.beforeAverageRating	Before-window average when ratings are available.
statistics.afterAverageRating	After-window average when ratings are available.
statistics.ratingChange	After-window average minus before-window average.
topIssues	Ranked app-level issue summaries.
topFeatureRequests	Ranked app-level feature-request summaries.
topPositiveTopics	Ranked positive topic summaries.
topNegativeTopics	Ranked negative topic summaries.
languageInsights	App-level language summary.
countryInsights	App-level country summary.
versionInsights	App-level version summary.
topicChanges	Observed before/after topic changes.
possibleRegressions	Topic increases worded as possible regressions, never causal proof.
release / windows	Release metadata and exact non-overlapping comparison windows.
release.version	Release version supplied to Release Impact.
release.releasedAt	Release timestamp supplied to Release Impact.
windows.before.from	Before-window ISO start.
windows.before.to	Before-window ISO end.
windows.after.from	After-window ISO start.
windows.after.to	After-window ISO end.
issueChanges	Issue signals across the release boundary.
featureRequestChanges	Feature-request signals across the release boundary.
newIssues	Issue signals first observed in the after window.
increasingIssues	Issue signals with increased after-window mentions.
decreasingIssues	Issue signals with decreased after-window mentions.
newFeatureRequests	Feature requests first observed in the after window.
compatibilityChanges	Compatibility-topic changes.
countryChanges	Country-dimension review volume changes.
languageChanges	Language-dimension review volume changes.
versionChanges	Version-dimension review volume changes.
warnings	Structured NO_REVIEWS, LIMITED_DATA, or future-release warnings.
generatedAt	Timestamp when a report or comparison was generated.
disclaimer	Caution that release comparisons are observational, not causal proof.

Aggregate reports are also stored under APP_REPORT_<app-id> in the default key-value store. Reports tolerate partial analysis failures by counting only successful analyses in ranked intelligence while retaining collection counts.

In releaseImpact mode, the enriched report is also stored under APP_RELEASE_IMPACT_<app-id>. Reviews exactly at release.releasedAt belong to the after window; reviews before it belong to the before window. Empty or small windows remain structured rather than being presented as proof of a regression.

The dataset schema provides thematic views for Reviews, Issue clusters, App reports, and Release impact reports. The views are projections over the same dataset; recordType remains the authoritative discriminator.

Issue-cluster output

feedbackCluster records contain a stable cluster ID, canonical issue or request title, feedback type, topics, mention counts, languages, countries, versions, severity estimate, confidence, and example/source review IDs. Clusters are partitioned by app and are emitted separately from raw reviews.

Aggregated app reports

productFeedbackReport records summarize collection volume, analyzed volume, ratings, actionable feedback, top issues, feature requests, positive/negative topics, and language/country/version insights. The same per-app report is stored as APP_REPORT_<app-id>.

Release Impact

releaseImpact mode compares non-overlapping review windows around a supplied release timestamp. It reports rating and volume changes, new/increasing/decreasing issue and feature-request signals, compatibility topics, and locale/version dimensions. It uses wording such as “possible regression” and “newly observed complaint”; timing alone is never treated as causal proof.

Language and country support

English and Vietnamese parsing are covered by redacted fixtures. Release Impact can request multiple language/country combinations, while every output keeps the request provenance. Review availability and date/label formatting depend on Google Play’s public response for the selected market.

Cost controls

Keep maxReviewsPerApp or maxReviewsPerPeriod bounded, request only required apps and locale combinations, and leave browser expansion disabled unless a later release explicitly supports it. The default deterministic fallback does not call an external model provider; external-provider and cloud billing costs are not included in the local benchmark.

Local run

npm install
apify validate-schema
apify run --purge --input-file sample-input.json

The local dataset and RUN_STATS key-value record are written under storage/. They are local verification artifacts and are not published automatically.

Benchmark evidence

The Phase 13 five-app matrix processed 15 reviews with 15 deterministic analyses, 0 collection errors, 0 analysis failures, 5 aggregate reports, 6.56 MB of public response data, 1.889 seconds runtime, and 200 MiB reported process RSS. See BENCHMARK_NOTES.md, docs/BENCHMARK_REPORT.md, docs/QUALITY_REVIEW.md, and docs/COST_REPORT.md for scope and limitations. These are operational measurements, not human-labeled accuracy claims.

Known limitations

• Google Play localizes markup labels and dates; parsing uses structural selectors and star classes rather than English-only labels.
• Direct HTML normally contains a small sample. “See all reviews” browser expansion is a later phase.
• App version and device metadata are nullable until a stable public fixture proves those fields.
• Public HTML is untrusted input; the Actor bounds requests, strips markup through Cheerio, validates ratings, deduplicates review IDs, and does not emit reviewer names or avatar URLs.
• The authenticated Google Play Developer API is a separate app-owner source and is not used for arbitrary public apps.

Responsible use

• Reviews are user opinions and may be incomplete, sarcastic, duplicated, or ambiguous.
• Detected issues are not confirmed bugs; severity and actionability are estimates.
• Device, operating-system, and app-version fields may be unavailable and remain nullable rather than inferred.
• Release Impact shows correlation around a date, not proven causation.
• Review important product decisions manually and respect Google’s terms, privacy expectations, and applicable law.

FAQ

Does this collect complete Google Play history? No. The current public HTML path returns a bounded server-rendered sample.

Does it create Jira/Linear tickets or use the Developer API? No. Those integrations are outside the first release, and the authenticated Developer API is not a public-app source.

Can I trust a cluster as a confirmed defect? No. Use clusters as prioritization evidence and trace them to the underlying reviews.

How do I read reports? Use the dataset views or read APP_REPORT_<app-id> and, for Release Impact, APP_RELEASE_IMPACT_<app-id> from the default key-value store.

What happens when a release window has too little data? The report remains structured and includes NO_REVIEWS or LIMITED_DATA warnings.

Search terms

Google Play reviews, Google Play review scraper, app review analyzer, mobile app feedback, app bug detector, feature request extractor, app sentiment analysis, release impact, Android app reviews, subscription complaints, app compatibility issues, mobile product feedback, app QA feedback, review clustering, Vietnamese app reviews.

Roadmap

Phase 16 runs the final publish-readiness matrix. Browser expansion, external-provider configuration, Apple App Store support, Reddit scraping, and automatic ticket creation remain deferred.