Frequently asked questions about deepsight cloud

deepsight cloud is a platform for surveys and for analyzing survey and feedback data: surveys, data import and connectors, analysis of open-ended and closed-ended responses, chat with your data, and dashboards and reporting.

Analysis features include topic analysis, sentiment analysis, NPS analysis, intent analysis, anonymization, translation and data quality checks. The cloud runs on Microsoft Azure in the EU.

Go to deepsight cloud →

Yes. Surveys are part of deepsight cloud: you create questionnaires, collect open-ended and closed-ended responses and analyze the results directly in the cloud. If you wish, the AI asks follow-up questions on open-ended responses in line with your research goal. Participation is currently in German.

Introductory customers receive the surveys package at no extra charge with a paid cloud plan. The duration and usage scope are specified in your offer before you start, and we activate surveys for your organization as part of the introduction. You receive a transparent follow-up offer before it ends; there is no automatic paid extension. See our pricing page for the terms.

View surveys →

deepsight cloud is suitable for, among others:

• Market research: open-ended responses, NPS comments, customer feedback
• HR departments: employee surveys, exit interviews, pulse surveys
• E-commerce: product reviews, customer reviews
• Organizations with large data volumes: separate organizations for departments or clients, SSO

No programming knowledge required – you work directly in the web app.

1. Request a demo: get to know the solution in a personal demo.
2. Receive test access: we agree the scope of the test with you, based on your project.
3. Upload your data as CSV or Excel and start the analysis.

If needed, we show you during or after the test how to work with your own data. No installation required.

deepsight is a specialized platform for systematic text analysis, while ChatGPT is a general-purpose chatbot. Key differences:

• Structured results: deepsight assigns topics and sentiment to every response and delivers tables, dashboards and exportable data; you check answers in the chat about your data against figures and evidence, too
• Reproducibility: saved results and topic versions can be reviewed again; new AI runs may differ
• Operation: Microsoft Azure in the EU; Microsoft does not use your data to train AI models
• Scalability: many texts in one run, not one at a time

Learn more about our methodology.

Topic analysis organizes open-ended responses by topic – with newly suggested categories or your own codebook.

Example (fictitious): in 2,000 NPS comments, topic analysis finds topics such as “customer service”, “product quality”, “delivery time” or “value for money”.

You get a matrix view (each topic as its own column) and can then refine the topics: rename, merge, split or remove them.

Sentiment analysis classifies the sentiment for each topic in a response – positive, neutral or negative – rather than only for the response as a whole. You check every classification against the original response.

Examples (fictitious):

• “Great product, fast delivery!” → product positive, delivery positive
• “Delivery on time, but the service was unfriendly.” → delivery positive, service negative

This shows which topics tend to be praised and which tend to be criticized. Learn more on the sentiment analysis page.

The deepsight cloud interface is currently in German.

Survey participation is currently in German.

You can also analyze responses in other languages. Integrated translation (AI-based, optionally also DeepL or Google) supports the analysis.

Recommendation: first translate the texts into one common language, then discover topics – this gives you consistent topics across all languages.

Yes. Use your own codebook – including parent and child categories – to analyze open-ended responses. You can assign responses strictly to your codebook or use it as a starting point and discover additional topics.

You upload your codebook as Excel (.xlsx) or CSV, optionally with a description and parent category for each category; both inform the assignment. If your codebook lists the codes as category names and the plain text as the description, you export either codes or descriptions.

Need a custom-trained model? As part of a custom project, we assess whether this approach suits your use case.

With a codebook, you define categories and their descriptions, for example for political topics in open-ended survey responses. Descriptions and parent categories inform the assignment.

Analysis profiles store study context, the analysis goal and a glossary with term meanings. When discovering and naming topics, the cloud takes this context into account, for example with specialist language or ambiguous comments. This aligns the analysis with your research question.

Automatic assignments can contain errors. Check them against the original responses and refine categories or context as needed.

Learn more about topic analysis and evidence and analysis profiles.

Discover new topics: deepsight suggests topics based on your texts, without predefined categories. Ideal when you don’t have a category system yet or want to discover new perspectives.

Apply your own codebook: deepsight assigns responses to your existing category system. Ideal for recurring studies with established categories.

You can combine both approaches: explore openly first, then build a codebook from the results. Learn more on our topic analysis page.

Yes. You can upload online feedback as CSV or Excel, for example:

• Google, Trustpilot, App Store and Google Play reviews
• Social media comments
• Forum posts and community feedback

For the Apple App Store and Google Play, every paid plan includes connectors that retrieve reviews regularly. They need access to your app: for Google Play, a service account and, for complete months, the Play Console reports bucket as well; without an App Store Connect key, deepsight only retrieves a limited look-back for one storefront country from the App Store. Several sources in one project run as separate time series side by side. Import other sources as CSV or Excel.

The combination of topic analysis and sentiment analysis shows which aspects are rated positively or negatively.

The recommendation question is answered on a scale from 0 to 10. The NPS summarizes the distribution: share of promoters minus share of detractors. It can range from −100 to +100.

NPS analysis links this score with the open-ended responses: you see which topics promoters, passives and detractors mention and check the comments behind every score.

Required: a text column and an NPS column (0–10). When you import data, deepsight suggests a suitable NPS column and you confirm the assignment. Learn more on the NPS analysis page.

Intent analysis classifies texts by the underlying intention: Is someone complaining? Is a feature being requested? Is someone expressing an intention to cancel?

The standard categories are complaint, praise, question, feature request, bug report and intention to cancel; you can extend them or replace them with your own. One text can express several intents. You check every classification against the passage in the text. Learn more on the intent analysis page.

Group comparison shows how topics and ratings differ between groups – for example by region, customer type, channel or department.

You choose suitable attributes from your data; deepsight suggests columns that fit. On request, deepsight flags statistically notable differences and weights the results – using a weight column from your data or to target distributions (rim weighting). Weights do not automatically make a sample representative. A difference shows where to look more closely – not why it occurs. Learn more on the group comparison page.

Shares and figures refer to the analyzed responses; anything the data quality check filters out does not count toward the base. One response can be assigned to several topics.

Fixed thresholds apply to small groups: below a base of 10 responses, deepsight flags values or hides them, and significance indicators require at least 30 responses per group. These thresholds cannot be changed.

For scales, deepsight shows the mean as well as top-2 and bottom-2 shares; you decide which end counts as good.

deepsight cloud is designed for data-protection-compliant use in organizations. Specifically:

• Runs on Microsoft Azure in the EU. For AI features we use Azure OpenAI, also in the EU; Microsoft does not use your data to train AI models.
• Data processing agreement (DPA) under Art. 28 GDPR. An overview of the service providers we use is available on request.
• Our own anonymization without a language model: personal data is detected and masked before the analyses run. If translation is switched on, texts are translated before masking. Analyses started directly from a survey run without anonymization. The anonymization does not detect every piece of information – review results before sharing them.
• ISO 27001 certification in preparation.

Find details on our security page.

Your data belongs to you:

• No training by Microsoft: for AI features we use Azure OpenAI; Microsoft does not use your data to train AI models.
• Encryption in transit (TLS) and at rest (AES-256).
• Data separation between customers.
• Retention periods you control, with advance notice.
• Support access to content only with time-limited approval from your admin.

Yes. HR data is particularly sensitive. That’s what anonymization is for – our own solution without a language model: personal data such as names, email addresses or phone numbers is detected and masked before the analyses run. If translation is switched on, texts are translated before masking.

Example (fictitious): “John Doe (john@company.com) is unhappy” → “[PERSON] ([EMAIL]) is unhappy”

The anonymization does not detect every piece of information – review results before sharing them. Analyses started directly from a survey run without anonymization. If needed, agree its use with your works council.

The anonymization feature is our own solution without a language model. It detects personal data such as names, email addresses and phone numbers and masks it before the analyses run (example: John Smith → [PERSON]). If translation is switched on, texts are translated before masking. Analyses started directly from a survey run without anonymization.

The anonymization does not detect every piece of information – review results before sharing them. You can also mask individual passages manually. Find details on our security page.

Support access to your content is only possible with time-limited approval from your admin. For a support case, you grant approval explicitly and for a limited period; access then expires automatically.

Admins manage users, roles and support approvals for their organization. Within an organization, all users see all projects.

In the Enterprise plan, we set up separate organizations for departments or clients that should not see each other. Each email address belongs to exactly one organization.

You can share individual results with external people, with a recipient list, password, expiry date and revocation. Guests have read-only access to these results.

Yes, in the Enterprise plan: SSO via OpenID Connect, tested with Microsoft Entra ID; other OpenID Connect providers on request. We set up the connection together with your IT team. If you wish, SSO sign-in can be made mandatory for your organization.

Learn more on the enterprise requirements page.

You decide. You set a retention period per project or for your organization; it counts from upload or analysis. deepsight warns you by email beforehand, and every deletion is recorded in the deletion log.

If you wish, only the original data is deleted; figures, dashboards and reports remain – including verbatim quotes you have added to reports. Raw responses from surveys in deepsight cloud are currently not covered by the retention period.

deepsight cloud is offered in plans with a fixed monthly price – Starter, Professional and Enterprise – from €299 per month. Every paid plan includes platform access with a text allowance, all analysis features including wave comparison and review connectors as well as reports, slides and PowerPoint export. The plans differ in volume (text allowance, users, assistant requests), in support and in Enterprise services.

Introductory customers receive the surveys package at no extra charge with a paid cloud plan. The duration and usage scope are specified in your offer before you start, and we activate surveys for your organization as part of the introduction. You receive a transparent follow-up offer before it ends; there is no automatic paid extension. See our pricing page for current prices and terms.

There is no free plan. Get to know the solution in a personal demo. Afterwards you receive test access. We agree the scope of the test with you, based on your project.

For regular use, choose one of the plans on our pricing page.

You pay a fixed plan price, monthly or annually. With annual billing you get 2 months free.

• Analyzed open-ended responses count toward your text allowance.
• If needed, you can add extra text packages – no plan change required.

See our pricing page for details.

Yes. The Enterprise plan (from €1,990 per month, from 50,000 texts per month) is designed for organizations with high analysis volumes. It includes:

• SSO via OpenID Connect (tested with Microsoft Entra ID)
• Multi-tenancy (separate organizations)
• On-premises on request
• Dedicated success manager
• Custom SLA

Contact us for a custom quote.

We currently do not offer an API for customers.

• Import: upload CSV or Excel directly in the web app; for selected sources also via connectors
• Export results as Excel or CSV
• We build special system integrations as custom projects

Talk to us if you need an automated integration.

Input formats:

• CSV (UTF-8, with or without header row)
• Excel (.xlsx); save older .xls files as .xlsx first

Output formats:

• CSV (with new columns for topics, sentiment, etc.)
• Excel (.xlsx)

For text analysis, the dataset needs at least one column with open-ended responses (e.g. “Comment”, “Feedback”, “Text”). Ratings and attributes come along in further columns. You import existing cross-tab reports in Excel format (common German-language layouts) separately from the raw data.

How many texts you analyze depends on the text allowance of your plan. If needed, you can add extra text packages – no plan change required.

For very large data volumes, we agree the approach with you individually in the Enterprise plan.

There is no real-time monitoring with alerts. When an analysis is finished, deepsight notifies you by email if you wish.

Yes, in several ways:

• Import and export: upload data as CSV or Excel, export results as Excel or CSV
• Connectors: recurring retrieval for selected sources
• Custom integration: we build special system integrations (e.g. with CRM or BI tools) as a custom project when needed

Learn more on the deepsight cloud page.

For all plans:

• Email support: send a support request
• Guides and use cases on our blog
• Feedback channel for feature requests

Professional and Enterprise: priority support.

Enterprise also includes:

• Dedicated success manager
• Custom SLA

Yes. Get to know the solution in a personal demo. Afterwards you receive test access. We agree the scope of the test with you, based on your project.

Request a demo →