Review Intelligence · the rating is the symptom

Your users already wrote the roadmap. It is sitting in your one-star reviews.

Nobody reads 500 reviews. So the complaint quietly killing trial-to-paid stays invisible until it surfaces as a rating drop, three releases too late — by which point it is an ASO problem, a paid-traffic problem, and a refund problem too.

This reads them properly: complaints grouped into themes, themes ranked by what they cost you, each one tagged to the thing that actually broke. You get a decision, not a sentiment score.

Try Review Retriever →See what comes back →

500Reviews per pull, read properly. Counting 50,000 tells you nothing you can act on

4Conditions every theme is tagged to — relevance, desire, trust, ability

2Stores. Yours or any competitor sitting on them

£0Free. No signup, no email capture, no seat you forget to cancel

The output

What comes back

A sentiment dashboard will tell you 62% of your reviews are negative, colour it red, and hand the problem straight back to you. You already knew it was bad. You clicked the rating.

The useful question is which complaints are the same complaint wearing different words, and which of those is large enough to be worth someone’s week. That is a judgement call, and it needs the reviews read — not counted.

Complaint themes

Grouped by what actually broke: price, trust, onboarding, bugs, value clarity, support, cancellation, expectation mismatch. Eleven scattered angry reviews usually turn out to be one billing problem.

Feature requests

Loud is not the same as valuable. Requests get separated by whether they are tied to use frequency, switching intent, or money — so the roadmap follows revenue instead of volume.

Growth risk

Which complaints are costing you conversion, retention, rating quality, and paid traffic efficiency. Named specifically, so you can argue for the fix instead of asking for it.

What to fix first

An order of work. The thing dashboards refuse to produce, because ranking means committing to being wrong in public.

One method · three routes

Start analyzing

Pick where you already work. One setup, then you paste a store URL and ask a question.

Claude

The skill plus a read-only connector. After setup you paste any App Store or Play URL straight into the chat — no export step, no file to keep track of.

Set up in Claude →

Codex

One plugin carries both the skill and the connector. Two commands in the terminal and you are pasting URLs a minute later.

Set up in Codex →

Custom GPT

Nothing to install. Export the reviews as Markdown, upload the file, ask. The slow route, and the one that works anywhere.

Set up the GPT →

Same method in all three. The only difference is how the reviews arrive. Full instructions, the configuration reference, and the data boundary live on the setup page.

One theme, fully resolved

What a finished theme looks like

Sentiment analysis stops at “negative sentiment increased.” That sentence has never once told anyone what to do on Monday. Here is where the work actually starts.

Theme

Cancellation trust gap

Users say the trial is easy to start and weirdly hard to stop. Twenty-three reviews, four different phrasings, one problem.

Growth impact

High

It poisons store-page trust, wastes paid traffic on users who read the reviews before installing, drags rating quality, and closes the door on subscription recovery. In that order.

Fix order

Before the next creative test

No screenshot compensates for users publicly warning each other that billing feels unsafe. Fix the flow, then spend on the funnel.

Evidence

Quoted, with IDs

Every theme cites the reviews it came from, so the product team can read the actual sentences instead of arguing with your summary.

A one-star about price, a one-star about crashes, and a one-star about feeling misled are three problems. Sentiment calls them one.

Sentiment tells you the review sounds negative. You knew that from the star rating, for free, instantly. What you are missing is the reason underneath: which expectation broke, which commitment got too heavy, and which fix moves money rather than moving a chart.

Roll all three into “negative sentiment” and you have built something very good at not naming anyone’s problem. Which is, quietly, why that format survives.

Coverage

Where this gets used

App Store

App Store review analysis

App Store reviews compress product, pricing, onboarding, and expectation failures into public language. The bad ones say the quiet part out loud: what users thought they were buying, what they got instead, and the exact sentence where trust cracked.

Run it when a release moves your ratings, when paid traffic starts delivering weaker users, when trial-to-paid slips, or when someone in the room says the problem is “just the App Store copy.” Sometimes it is. More often the store page is only where the damage becomes visible.

Google Play

Google Play review analysis

Play reviews are where the messy operational failures live: device-specific bugs, country patterns, subscription confusion, refund anger, login loops, support gaps, feature expectations that split by segment. Users are blunter here, and they name the handset.

Three questions get answered. Which complaint cluster is big enough to matter, which one is getting worse, and which one is blocking the next commitment you need from the user.

Competitors

Competitor app review analysis

Your competitor’s reviews are public and they are free. Every recurring one-star theme is a user segment they are failing, described in that user’s own words — which makes it a positioning brief you did not have to run a study to get.

Use it before changing ASO creative, building a comparison page, repositioning a feature, or deciding which promise is safe to make in paid ads. Their review section will tell you exactly which promise they cannot cash. Make that one.

What it does not do

It does not get you reviews, write reviews, or lift your rating. If you arrived here looking for a way to buy reviews: wrong page, and probably wrong strategy.

It also does not replace talking to users. It tells you which five conversations are worth booking, which is the part most research plans get wrong before they start.

Four things people ask first

FAQ

Do I need this? I read my reviews.

You read the last twenty, sorted by most recent, on a Tuesday, in a bad mood. That is not the same job as 500 sorted, grouped, and ranked.

The value is not in reading. It is in noticing that eleven separate complaints about eleven different screens are one billing problem — which is exactly the pattern that survives casual reading.

Can I run a competitor’s reviews?

Yes, and it is usually the higher-value run. Their reviews are public demand research nobody is paying for. The question is not whether their users are annoyed — it is which of their complaint patterns opens a gap for your product, your positioning, or your next set of creative.

Is this just sentiment analysis with better branding?

No. Sentiment is a finish line for people who do not want to name the problem. This groups complaints by theme, severity, likely revenue impact, and the broken commitment — relevance, desire, trust, or ability — and then puts them in an order.

Free how, exactly?

Free as in no signup, no email capture, no trial that quietly becomes a subscription. Public reviews are public; retrieving and reading them costs almost nothing.

The paid work is further down the page, and it is a different thing: a decision memo for teams with real money riding on the answer. The tool is useful on its own and you are welcome to never speak to me.

When this becomes a paid problem

If the analysis is feeding a real growth decision, “top complaints” is not the output. A memo is.

Send me the ugly version →

Which competitor weakness to attack. Which store-page promise to change. Which product gap is actually blocking money rather than merely annoying people. And which claim to avoid entirely, because the product cannot prove it yet and the reviews will say so within a quarter.

For teams with enough review volume, subscription risk, or paid traffic at stake, that belongs in a direct conversation — not a research task that gets admired and filed.

Subscription version starts with trial · pricing · cancellation · renewal · value-gap evidence

Run your reviews. Or your competitor’s.

Free, no signup. Worst case you confirm what you suspected and lose an afternoon.

Start analyzing →Send me the ugly version →