Solutions · CX & Voice of Customer Organize customer feedback by topic and compare it.
Compare which topics come up over time and alongside different ratings. The original comments help you interpret the differences.
Your customers say a lot – just not in one place.
And new comments arrive every month. Turning them into something service, product and sales teams can work with takes a lot of manual effort.
The challenge: Every channel on its own
Survey tool, app stores and ticketing system each produce their own reports with their own categories. Nobody sees whether the same criticism shows up everywhere.
With deepsight: The same categories for every source
Analyze every source with the same codebook, including parent and child categories. You see which channels a topic shows up in.
About your own codebookThe challenge: A score without substance
The report shows the NPS and perhaps a word cloud. Neither tells you which topics detractors raise.
With deepsight: Topics by rating group
deepsight sorts the comments by topic and compares what detractors, passives and promoters mention – with the comments verbatim.
Go to NPS analysisThe challenge: Starting from scratch every month
New export, new count. If the categories come out slightly differently each time, it stays unclear whether a topic is really growing.
With deepsight: A time series instead of one-off reports
Add new exports and newly retrieved app reviews as a further wave and align the topics. You see what increases and what decreases.
Go to time seriesThe challenge: Individual requests get lost
Among hundreds of comments are complaints, open questions and explicitly announced cancellations. To find them, someone has to read everything.
With deepsight: Filter by intent
deepsight sorts comments by intent, such as a complaint, a question or an explicitly stated intention to cancel. Filter these comments and pass them on to your service team.
Go to intent analysis
Same topics, two months
Comments from a customer survey, exported as a CSV file. Select a topic to see a comment from August.
Illustrative values. Share of comments that mention a topic. Base: July n = 812, August n = 764.
“On the hotline I was stuck on hold for half an hour, but after that it was quick.”
Select a category to see the matching passage.
How your feedback gets into deepsight
| Source | How it gets into deepsight | Status |
|---|---|---|
| Survey and feedback exports | Import as an Excel or CSV file, including ratings and attributes. Compare recurring exports as waves over time (included in every paid plan, see the plan comparison). | Available |
| Your own surveys | Create a questionnaire with surveys in deepsight cloud and analyze the responses right away (activated during onboarding; participation currently in German). | Available |
| Review platforms | App Store and Google Play via a connector that retrieves reviews regularly (included in every paid plan); which data a source provides depends on access and permissions. Import other platforms such as Google or Trustpilot as an export. | Available |
| Support tickets | Import an export from your ticketing system as an Excel or CSV file. | Available |
| Connecting other systems directly | For example your ticketing system: as a custom project, after a joint assessment. | On request |
What your customer feedback can tell you
- Which topics do customers mention more often this month than last month – and which less often?
- Which topics are praised, which are criticized – and does a topic tip from praise to criticism?
- Do new topics appear after a change, such as a new pricing plan or an app update?
- Do regions, products or customer groups differ in what they write about?
- For which topics do customers explicitly state an intention to cancel?
- Which questions do customers ask again and again, for example about billing or contracts?
Good to know: You can also ask the assistant questions like these in your own words. Its answer separates calculation, evidence and interpretation.
What your export should contain
One row per comment. What sits next to it determines what you can compare later.
Good to know: Tickets and comments often contain names or contact details. Anonymization masks details such as names, email addresses or phone numbers before the analyses run. It does not detect every piece of information – check results before sharing them.
A single comment is an anecdote. Sorted by topic, comments show what keeps coming up.
Across all sources, month after month.
From export to topic comparison
- Bring in the data
Upload exports or retrieve app reviews via connector. Filter out empty comments and entries like “no comment” if you wish.
Data quality check - Set the topics
Let deepsight suggest topics or apply your codebook. Add product names, plans and internal terms to the study context beforehand.
Topic analysis - Check against the comments
Read the comments for each topic verbatim, correct individual assignments, and rename, merge or split topics.
Topic refinement - Compare
Put ratings and topics side by side by month, source or customer group, with significance indicators if you wish.
Group comparison
What you pass on to service, product and sales
Good to know: Your quarterly report is built from tables produced in a different program? The slide factory turns Excel data into slides based on your PowerPoint template.
When service, product and sales see the same topics, everyone talks about the same thing.
Instead of a separate report for each channel.
What matters when you compare feedback
Add new topics deliberately
Apply your codebook with additional suggested topics, and you’ll notice what it doesn’t cover yet – after an app update, for example. You decide whether a new topic stays: keep, merge or remove it. Every change is saved as a topic version.
Base per month and source
Shares refer to the analyzed comments of one month or one source; anything the data quality check filters out doesn’t count toward the base. Multiple mentions are possible. With few comments, shares fluctuate strongly – below a base of 10, values are flagged or hidden in group comparison and dashboards.
Every source has its own audience
Customers write tickets when they need something; reviews often come from customers with particularly good or bad experiences. What dominates tickets therefore doesn’t automatically apply to all customers.
Sentiment by topic
A comment like “Great product, late delivery” praises and criticizes at the same time. deepsight therefore classifies sentiment for each topic; check irony and ambiguous wording against the original.
Questions about CX & Voice of Customer
As often as new data comes in, for example monthly or per survey wave. Analyzed comments count toward your plan’s text allowance, see Pricing.
There is no real-time monitoring with alerts: you see changes when you analyze and compare new data. When an analysis is finished, deepsight can notify you by email.
Yes, if you analyze the earlier exports with the same categories as the new ones, for example using your existing category system as a codebook. We don’t recommend placing shares from the old tool directly alongside: different categories and assignment rules shift the figures. Earlier exports also count toward your text allowance when you analyze them. Wave comparison is included in every paid plan – see the plan comparison.
Yes. An integrated translation – AI-powered, or with DeepL or Google if you prefer – supports the analysis, and the original comments remain available. If translation is switched on, texts are translated before anonymization masks personal information.
Not in the sense of a cause. Frequently mentioned topics show where to take a closer look; significance indicators show whether a group differs noticeably from the rest of the sample in statistical terms. Neither proves that a topic changes the score.
deepsight cloud runs on Microsoft Azure in the EU. Data is encrypted in transit (TLS) and at rest (AES-256), and you set retention periods for datasets per project or for the whole organization. For AI features we use Azure OpenAI; Microsoft does not use your data to train AI models. A data processing agreement under Art. 28 GDPR governs the processing. More under Security.
Our anonymization masks personal information such as names, email addresses or phone numbers before the analyses run; if translation is switched on, texts are translated first. It does not detect every piece of information.
Related to CX & VoC
Analyze your customer feedback
See the topic comparison in a personal demo. Afterwards, you’ll receive test access – and if you like, we’ll go through the analysis using an export from your system.