Google Play Reviews Scraper & App Intelligence

Turn public Google Play reviews into structured product-feedback evidence. Instead of manually reading Store pages, provide Android package IDs or public app URLs and receive review rows, app summaries, OUTPUT, and REPORT.html for product, QA, ASO, and reputation-review decisions.

What problem it solves

Ratings alone do not explain what users are reporting, whether feedback clusters around a version, or how often developers answer. This Actor collects the public evidence and summarizes the observed sample with transparent, deterministic calculations. It does not estimate downloads, revenue, conversion, sentiment, or future ratings.

Input → result → decision

1. Input: one or more Android package IDs or Google Play URLs, plus country, language, ordering, star filters, and a bounded review limit.
2. Result: deduplicated public review rows, one non-billable app summary per app, aggregate OUTPUT, and a self-contained HTML report.
3. Decision: identify feedback that merits manual investigation, compare observed app-health signals, or prioritize QA and response workflows.

Three concrete use cases

• Product feedback review: surface recurring and complaint terms in the collected sample before reading individual reviews.
• QA and release monitoring: inspect version issue signals and recent-versus-previous rating movement as descriptive evidence, never as causation.
• Reputation response planning: review public developer-reply coverage and filter low-star feedback for a response queue.

Quick Start

Run a small, bounded collection first:

{
  "apps": ["com.spotify.music"],
  "maxReviewsPerApp": 25,
  "sortBy": "newest",
  "language": "en",
  "country": "us",
  "includeDeveloperReplies": true,
  "includeAppDetails": true
}

Then inspect the reviews and app_summaries Dataset views. Use OUTPUT for aggregate counts and billing metadata, or REPORT.html for a readable evidence review. A reusable input is available at examples/STORE_DEMO_INPUT.json.

Sanitized output example

{
  "record_type": "review",
  "app_id": "com.example.app",
  "review_id": "fixture-review-001",
  "rating": 2,
  "review_text": "slow login after update",
  "app_version": "9.1.0",
  "developer_reply": "Thanks for reporting",
  "source": "sanitized deterministic fixture"
}

The complete local example is examples/STORE_DEMO_OUTPUT.json. It was derived from deterministic test-fixture shapes and anonymized; it is not a live Cloud run, external production result, reliability claim, or proof of future availability.

Main fields

Review rows include:

• app_id, app_url, and app_title
• review_id, rating, review_text, and review_created_at
• thumbs_up_count, app_version, and public developer-reply fields
• language, country, source, and retrieved_at

Non-billable app summaries can include:

• observed rating distribution and positive/negative share
• recent and previous observed averages plus descriptive trend
• developer-response coverage
• recurring terms, complaint terms, and version issue signals
• explainable App Health and Product Feedback Opportunity scores with confidence and reasons

Unknown or unexposed values remain null rather than being invented.

Clear pricing

The active Cloud event is review-result at USD 0.0002 per valid unique review delivered.

Delivered reviews	Event cost
100	USD 0.02
1,000	USD 0.20
10,000	USD 2.00

App summaries, duplicates, invalid or empty review rows, warnings, OUTPUT, and REPORT.html are non-billable. Platform compute or storage charges, when applicable to the account, are separate from these event-cost examples.

Verifiable proof points

• Billing eligibility requires a non-empty valid review row and is evaluated after Dataset delivery.
• Review delivery is deduplicated by review_id, with a documented content fallback when the public identifier is absent.
• Store rating and the rating observed in the collected review sample remain separate fields.
• Scores expose their component evidence and do not use generative sentiment or synthetic values.
• Dataset, OUTPUT, and REPORT.html preserve provenance, warnings, and null handling.

Source and method

The Actor uses anonymous HTTPS requests to the public Google Play web surface: the app details page and the web client's UsvDTd batchexecute review RPC. It does not require a Google API key, Play Console login, browser automation, or synthetic data.

App Health is a 0–100 explainable score based on observed rating quality, negative share, recent trend, developer-response coverage, and evidence coverage. Levels are HIGH (75+), MEDIUM (50–74.9), and LOW (<50). Confidence reflects observed sample size. These are descriptive summaries of collected public evidence, not predictions.

Honest limitations

• Google Play's public HTML and undocumented web RPC can change, throttle, or omit fields.
• Review availability and ordering vary by country, language, time, and Google's public response.
• A bounded sample is not the complete review population and must not be treated as market share or a reliability metric.
• Trend and version signals are correlations in observed rows; they do not establish causation.
• Term frequency is deterministic text counting, not semantic sentiment analysis.
• Public reviewer names are kept only when exposed and are not used for profiling.

Next step

Start with one app and 25 reviews, inspect the evidence quality and warnings, then increase the bound only if the returned sample supports the decision you need to make.

Related App Intelligence Actors

• Google Play Apps Scraper & ASO Market Intelligence
• Apple App Store Reviews Scraper & Rating Intelligence