OneStar is review analysis for apps and games. It reads App Store, Google Play and Steam reviews one at a time, counts what people love, why they leave and what they ask for, and links every number to the reviews behind it. So you can fix what hurts, protect what people love and answer the people behind the stars.
Natural mix · 2024-03-01 to 2026-09-25 · 6,000 reviewsNatural mix · 2024-03-01 to 2026-09-25 · 6,000
Top praise: works reliably and easy to use. Top complaint: crashes and bugs (in 49% of complaints), then bad update or redesign.
A popular app collects thousands of reviews a month, and the store boils them down to one number. That number moves, but it never says what moved it: the bug that came with the last update, the paywall that pushed out long time users or the feature a whole audience keeps asking for.
The reasons are all in the text. Reading it well, at scale, is the hard part.
Every report reads the same way, for any app or game, so findings compare across products and over time. The panels below come from a live public report: Has the Sonos app recovered from the 2024 redesign?, 6,000 reviews of Sonos from App Store and Google Play.
Complaint themes ranked by share, with how severe each one is and how often it ends with someone quitting.
The app no longer has any stability.
Praise themes ranked the same way, so a fix never breaks the thing people stay for.
I had originally rated this app as really poor but in fairness I think it must’ve been updated and the reliability and stability of the app is much much improved. Well done.
Feature requests sorted by type and by theme, so a roadmap starts from what people actually wrote.
13%of reviews ask for something new or for something back
New users against veterans, store against store, one star against five. A difference is called out only when it survives a significance test.
Weekly and monthly trends, the impact of each release, and spikes or review bombs flagged by code.
When Jev is not confident about a theme, the review goes to an unsure bucket that stays on screen instead of being forced into a category.
Free to read. A free account opens every section except the version table.
Search the App Store, Google Play or Steam by name. The listings of one product on several stores are matched into one report.
A version table shows how each release moved complaints and ratings, so a bad update is caught by its numbers, not a hunch.
Days with unusual review volume are flagged by code, so a coordinated wave is told apart from a real problem.
Write a statement such as "the price rise drove people away" and see how many reviews agree, disagree or never mention it, with quotes.
Two to four apps read on the same themes, side by side. A leader is named only when the gap holds up.
Daily or weekly refreshes, and an email when a complaint spikes, a rating drops or a review bomb starts.
Connect App Store Connect or Play Console (in beta) to read every country, the full history and your replies.
Add themes of your own, or ask what reviewers use the product for and which competitors they bring up.
A read only link for the team, a public page with its own preview image, or every judged review as a CSV file.
Four steps, the same for every report and every app. Nothing the model could get wrong is left to it when code can count it exactly.
App Store and Google Play reviews come through a review data service, or straight from App Store Connect and Play Console when you connect your own app. Steam reviews come from Steam's public API. Author names are never read and personal details are removed before anything else sees the text.
Jev, a model from TypeSafe built for typed judgments, answers the same 12 questions of every review. Each answer is one option from a fixed list with a confidence score, never free text, which is what lets code count it.
Plain code turns stored answers into shares, trends and audience splits. Each share carries a likely range, and an audience difference is called out only when it passes a significance test corrected for how many were tried. Moving a filter recounts stored answers and never asks the model again.
Every chart opens the reviews it was counted from, with Jev's answers beside each one, so anyone can check the work before acting on it.
The app no longer has any stability.
A language model labels each review. Everything else, from counting to dates to percentages, is plain code, so every figure is exact for the reviews read and can be rebuilt from stored answers.
When Jev is not confident, the review lands in an unsure bucket that stays on screen. Other and unclear stay visible too. A report that hides its doubt is not worth trusting.
Each question is scored against hand labeled sets of real reviews, 300 per kind of product, and the cutoff for when an answer counts as sure comes from those scores.
Author names and profiles are never read, personal details are removed before analysis and at most 200 characters of any review are kept.
Both work for a handful of reviews. Past a few hundred, this is how the options compare.
| What matters | Reading by hand | A general chatbot | OneStar |
|---|---|---|---|
| Reads every review in the sample | PartlyUntil time runs out | PartlyWhatever fits in one prompt | YesEach review on its own |
| Asks the same questions every time | PartlyDrifts as you tire | NoAnswers shift between runs | YesTwelve fixed questions, scored on labeled sets |
| Numbers you can reproduce | PartlyIf you kept a tally | NoCounts are the model's estimate | YesCounted by code from stored answers |
| Says what it is unsure of | NoRarely written down | NoSounds sure either way | YesUnsure answers stay on screen |
| Quotes you can check | YesYou read them yourself | PartlyNot checked against the source | YesEvery number opens its reviews |
| Rivals, three stores, over time | NoStart over each time | PartlyCollect and paste each time | YesSearch any app, track it on Pro |
OneStar also needs no sales call and no setup, is priced for one person and covers Steam games as well as apps.
For a product of your own or one you are studying. Each of these starts with a question the star rating cannot answer.
“Why did our rating drop after the last release?”
See the complaints that started with a version, in reviewers' own words, and know what to fix before the next one ships.
Search your app“Is this a review bomb or a real problem?”
Steam reviews read alongside days of unusual volume, so a patch backlash and a coordinated wave look different.
Steam reports“What should we fix first, and for whom?”
Rank complaints by share, severity and churn, protect what people praise and see which audiences want different things.
Browse public reports“Where are competitors losing their users?”
Study a market before building in it: where rivals bleed users, what their fans love and which requests nobody answers.
Most complained about“What are people upset about this week?”
Catch a new complaint or a review bomb the week it starts, with the quotes you need to answer people.
Biggest rating dropsEvery public report is free to read. These are among the most read this month.
Natural mix · 2025-06-01 to 2026-09-25 · 2,117 reviewsNatural mix · 2025-06-01 to 2026-09-25 · 2,117
Top praise: content variety and depth and easy to use. Top complaint: ads (in 34% of complaints), then pricing and paywall.
Also on Spotify: Music and Podcasts: Spotify vs Apple Music vs YouTube Music: what listeners complain about ·
Natural mix · last year · 1,894 reviews · refreshed weeklyNatural mix · last year · 1,894 · weekly
Top praise: gameplay is fun and multiplayer and community. Top complaint: cheaters and toxicity (in 67% of complaints).
Natural mix · 2025-10-01 to 2026-09-25 · 2,118 reviewsNatural mix · 2025-10-01 to 2026-09-25 · 2,118
Top praise: easy to use and works reliably. Top complaint: content quality (in 37% of complaints), then company, ethics and policy.
Priced for one person, not a procurement team. Public reports stay free to read either way.
$0no card needed
$19a month, or $190 a year
The same rules apply to a free report and a paid one.
Reviews are where OneStar starts listening, because they are public, honest and tied to a product. The goal is bigger: better tools for understanding and supporting audiences, from hearing them, to answering them, to following them wherever they talk.
OneStar is a review analysis tool for apps and games. It reads public reviews from the App Store, Google Play and Steam with the Jev model and reports what people love, why they leave, what they ask for, who is saying it and how that changes over time, with the reviews behind every number.
Indie developers, game studios, product managers, researchers, marketers and support teams who want to understand an audience, whether the product is their own or a competitor's.
A chatbot reads what fits in one prompt and estimates its numbers. OneStar asks the same 12 questions of every review on its own, stores each answer and counts them with code, so the numbers are exact for the reviews read, come out the same every time and link to the reviews they were counted from. Answers the model is unsure of are shown, not guessed.
Jev is a model made by TypeSafe for typed judgments. It answers each question with one option from a fixed list and a confidence score instead of free text, which is what lets plain code count its answers and set aside the ones it is unsure of.
The Apple App Store, Google Play and Steam. One report can combine every store a product ships on. The App Store gives at most the newest 500 reviews in each country a report reads, which on a busy app can cover only recent days or weeks, and tracked reports build a longer history with each refresh. Google Play reads back through up to about 30,000 of the newest reviews per report. On Pro and Max you can connect App Store Connect or Play Console (in beta), and your own app is then read through your account: every country, its history (on Google Play, from the report files Play Console keeps for you) and your replies.
Yes. Any app on the App Store or Google Play and any game on Steam can be searched and analyzed, because reports read public reviews and need nothing from the developer. On Pro, Compare puts two to four apps side by side on the same themes.
Public reports are free to read without an account. A free account opens every section of any shared report except the version table and gets 3 report runs a month, reruns included, of up to 500 reviews each, on apps you choose, no card needed. Pro, at $19 a month, adds the full analysis of your own reports, Ask the reviews, comparisons, tracking with alerts, CSV export and connections to your own App Store Connect and Play Console.
Each question is scored against hand labeled review sets before it ships, and the confidence an answer needs to count comes from those scores. Answers below it go to a visible unsure bucket instead of being forced into a category. Shares carry a likely range, and every number links to the reviews it was counted from.
No. Every review in a report is read, from one star to five. Reports show praise and feature requests next to complaints, split by audience and tracked over time.
OneStar never reads author names or profiles, removes personal details before analysis and stores at most a 200 character excerpt of any review.
Yes. Public reports can be quoted with a link to the report page. AI assistants and search tools can read a plain text index of every public report at onestar.sh/llms.txt.
Search any app or game. Public reports open in seconds, and a free account runs reports of your own, no card needed.
