App Review Complaint Benchmarks — Category Baselines + App Complaint Profiles

What you get

• S — Google Play only: category complaint baselines + app profiles (5,368 rows)
• M — Both stores: all category baselines + app complaint profiles (5,639 rows)
• L — Full library incl. per-topic top-50 worst-offender leaderboards (6,039 rows)
• Data dictionary, datasheet and license in every package

Coverage statistics

Records

6,039

Largest package

1.1 MB

Available packages

Real contents of the current builds — row counts, sizes and checksums come from the packaging manifest, not estimates.

Free Open Data — full downloads

This dataset is part of the DataForge Open Data program: the upstream license (CC BY-NC 4.0 (DataForge curated layer; academic/personal use, attribution + backlink required) — commercial use requires a DataForge commercial license) permits free redistribution, so the complete packages below are free — no signup. Please cite DataForge (data.zalize.com) with a backlink when you use or share this data.

License: CC BY-NC 4.0 (DataForge curated layer; academic/personal use, attribution + backlink required) — commercial use requires a DataForge commercial license · Attribution: DataForge (data.zalize.com)

Full package inventory

Every package registered in the delivery catalog for this dataset — 3 packages, 3.1 MB total , 17,046 rows. Sizes, row counts and SHA-256 checksums come straight from the packaging pipeline.

How to get this data

All packages are hosted on the Hugging Face Hub: free/open packages live in a public dataset repo (direct downloads, no signup); paid or restricted editions are delivered from a private repo with token access after approval.

# Download the whole dataset repo (hf CLI — pip install -U huggingface_hub)
hf download zalizedata/app-review-complaint-benchmarks --repo-type dataset --local-dir app-review-complaint-benchmarks

# Or just one archive
hf download zalizedata/app-review-complaint-benchmarks archive/complaint-benchmarks-tier1-L-2026-08-02.zip --repo-type dataset

Protocol access (hf://, resolve links, token auth for private repos) is documented in the developer docs →

Pricing

AI-training use requires the AI-Training tier (1.5× base); redistribution licensing is quoted individually. Annual subscriptions get 2 months free. License details →

Update frequency & delivery

• Computed from the DataForge appstore pipeline's real review collection run (no third-party scraping services).
• Complaint topics are classified from review text; the packages ship aggregated metrics only, no review full text.

Compliance & provenance

Collected only from publicly accessible pages and endpoints — no login-required data, no CAPTCHA solving, no forged authentication. Collection is rate-limited with an identified User-Agent, and collection logs are retained. PII is removed or pseudonymized. Every record carries its source URL and collection timestamp. Sold under a tiered commercial license.

FAQ