Solutions · Reviews & feedback Analyze reviews and compare them over time.
Import existing reviews or retrieve app reviews from the App Store and Google Play regularly via connector. Compare topics across time periods and check the comments behind them.
Two stars after the update. The text says why.
New reviews come in every month, often hundreds of them. Reading, sorting, comparing with the previous month: mostly manual work.
The challenge: Stars without reasons
The average drops after an update. Finding the topics behind it means reading one review after another.
With deepsight: Topics and sentiment behind the stars
deepsight sorts every review with text by topic and shows for each topic whether it praises or criticizes. With the star rating as an attribute, you see what one- and two-star reviews mention.
Go to sentiment analysisThe challenge: Export after export
Collect reviews from App Store Connect and the Play Console, copy them into a spreadsheet, sort them again. Next month, start over.
With deepsight: Retrieval via connector
A connector fetches your app’s reviews from the App Store and Google Play, with no manual export. Reviews from other portals can be uploaded as a file.
Go to data sourcesThe challenge: Months that don’t match up
Every new analysis builds its own categories. Whether a complaint becomes less frequent after the fix is hard to tell.
With deepsight: Topic matching across months
In wave comparison, deepsight matches the topics of each new month with the existing ones and suggests assignments, which you confirm.
Go to data and time seriesThe challenge: Bug reports mixed in with praise and requests
Bug reports, feature requests and questions sit in the same reviews. Product and support teams pick out their cases by hand.
With deepsight: Filtered by intent
Intent analysis classifies reviews as a bug report, feature request or question, for example – several in one review if needed. Each team filters for its own cases.
Go to intent analysis
One source, three months
- Source
- Google Play, “Sample Shop” app
- Retrieved
- Connector with Play Console report storage, as of Sep 1, 2026
- Base per month
- June 1,180 · July 1,090 · August 1,240
Illustrative values. Share of reviews in the month that mention a topic.
“Since the last update I have to log in again every time. Annoying.”
Select a category to see the matching passage.
Read channels side by side
People who review in the App Store are often a different audience from the participants in your customer survey, so different topics per source are to be expected. Compare sources side by side instead of merging them into a single score.
Competitor comparisons and industry benchmarks require you to bring the corresponding reviews. deepsight analyzes the data you upload or connect.
| Source | How it gets into deepsight | Status |
|---|---|---|
| Exports from review portals | For example Google, Trustpilot or Kununu: import as an Excel or CSV file. | Available |
| Google Play and App Store | A connector retrieves an app’s reviews regularly; each full calendar month is added as a further wave. The connector is included in every paid plan, see the plan comparison. Which data a source provides depends on access and permissions. | Available |
| Other platforms | Direct connection as a custom project, after a joint assessment. | On request |
From the monthly comparison to the product meeting
What you find in the reviews can be shared straight from deepsight cloud.
Dashboard
Put the monthly comparison and topics together in a dashboard – for your team or via a shared link.
Report and PowerPoint
Pin findings, share them as a report via a link or export them to PowerPoint. More on reporting.
Next month
Add newly fetched reviews to a time series; when an analysis is finished, deepsight can notify you by email. The connector and wave comparison are included in every paid plan – see the plan comparison.
What your reviews can tell you
- Which criticism appears after an update – and since which month or app version (Google Play)?
- Which bugs do users describe, and do they cluster on certain devices or in certain language regions?
- Which features do users ask for, and how do they describe what’s missing?
- What do satisfied users praise – and what should a redesign therefore keep?
- Which topics do users mention in the App Store, and which on Google Play – read side by side in two time series?
- Which reviews are about service, delivery or prices rather than the app itself?
Good to know: You can also ask the assistant questions like these directly. Depending on the question, it answers with charts and verbatim reviews as evidence.
What your data should include
Besides the text, the details of each review matter – use them as attributes in comparisons.
“Great app, but since the update it keeps crashing.” Five stars – and a bug report.
That’s why deepsight reads topics, sentiment and intent from the text – not from the stars.
From review to finding
- Bring in the data
Set up the connector or upload a portal export, and add app and feature names as key terms to the study context. Filter out entries without substance up front if you wish.
Data quality check - Build topics
deepsight suggests categories based on the reviews. In wave comparison, it analyzes each new month with the same settings as the previous one, such as the same codebook and translation.
Topic analysis - Check against the original
Read the reviews for each topic verbatim, correct assignments, merge or split categories. Every change is saved as a version.
Topic refinement - Compare
Put topics side by side by the details your source provides, such as star rating, month, app version or language, with significance indicators if you wish.
Group comparison
Good to know: Names, email addresses or phone numbers in reviews are masked by anonymization before the analyses run. If translation is switched on, texts are translated before masking. Anonymization doesn’t catch every detail – check the results before sharing them.
What you take away from the analysis
Behind every bug report in the chart is the review, word for word.
Your team discusses what users actually write, not the star average.
What matters with reviews
The text is the base
Star ratings without a comment don’t feed into the topics, so the average rating in the store may differ. Shares refer to the analyzed reviews in a month; anything the data quality check filters out doesn’t count toward the base. Below a base of 10, values are flagged or hidden in group comparison and dashboards.
Timing is not proof
If a topic rises after an update, a connection seems likely – but that doesn’t prove it. Compare by month or, where available, by app version and read the reviews behind it. Significance indicators show whether a group differs noticeably from the rest in statistical terms, not why.
Check irony against the original
“Great, logged out again” is criticism. Sentiment analysis doesn’t always recognize irony, so read conspicuous cases against the original during your review. The star rating helps: “Oh great” with one star is hardly meant as praise.
Questions about reviews & feedback
For regular retrieval, the connector needs access to your app: an App Store Connect key for Apple, and for Google a service account for the Play Console plus, for automatic monthly waves, the Play Console report storage (bucket). For Apple, the look-back at setup is limited and covers one store country; without a key, this limit applies to all retrieval.
Besides text, star rating and date, the connector provides the country (App Store) or language (Google Play); app version and device mainly for Google Play. It doesn’t collect reviewers’ names.
The connector groups reviews by calendar month, and each full month becomes a wave of its own. You set up the first wave once; after that, deepsight analyzes each further month automatically and, if you wish, notifies you by email when the analysis is finished. If the text allowance is used up, the source is visibly paused. The connector, time series and topic matching (wave comparison) are included in every paid plan – see the plan comparison.
deepsight is therefore designed for comparing periods, not for watching individual reviews in real time; there are no alerts for individual reviews. Until you have matched the topics of a new wave with the previous months, it counts as provisional.
Each connector is a separate source with its own time series; there is no combined time series for both stores. You read the topics from both stores side by side. To compare both stores as groups in one analysis, upload the reviews from both stores as one file with a column for the platform – that’s a separate analysis and counts toward your text allowance.
Each analyzed review counts as one text, whether it was retrieved via connector or uploaded as a file. A new completed analysis run can count the same texts again. Once the data quality check has been completed, filtered-out rows do not count; without this check, the dataset row count is used. The pricing page shows how many texts each plan includes.
Yes. Use your own codebook – including parent and child categories – for the analysis, sticking strictly to your categories or with additional suggested topics. No training required. Reviews and tickets can then be read against the same categories.
Related to reviews & feedback
Analyze your reviews
See the comparison over time in a personal demo, using a source from your field. Afterwards, you’ll receive test access.