Use cases

What does each use case measure?

When you pick a use case, which analyses run, which aspects get measured and what the report will look like are all known up front. Every one of them is written out below.

Product feedbackE-commerce feedbackNPS survey analysisOpen-ended survey analysisSocial feedback analysisMobile game feedbackRestaurant & hotelBanking & fintechChat & call centreEmployee feedback

Product feedback

Uncover needs, likes and complaints from product and app reviews.

Data

Pull Google Play and App Store reviews directly, or upload CSV/Excel. If a date column exists, trend is calculated too.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Aspect-based sentiment
  • Topic modeling
  • Keyword extraction
  • NPS/CSAT prediction
  • All four insights

Aspects measured

usabilityperformancedesignpricesupport

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Which aspect of the product is loved and which generates complaints; ranked pain points and an NPS estimate. Every finding backed by real quotes.

E-commerce feedback

See where the delivery chain and the seller experience break down.

Data

Marketplace review export (CSV/Excel). The review column is required; with a date column monthly trend is added.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Aspect-based sentiment
  • NPS/CSAT prediction
  • Keyword extraction
  • Trend analysis
  • All four insights

Aspects measured

shippingpackagingsellerpriceproduct qualityreturns

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Separate sentiment breakdown for shipping, packaging and returns. It separates whether the complaint is about the product or the delivery. That is the distinction a single sentiment score hides.

NPS survey analysis

The text says what the score does not. Compare the two.

Data

A survey with both a numeric score column and open-ended answers. Scores can be 0-10 or 1-5; the scale is normalised automatically.

300 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • NPS/CSAT predictionrequired
  • Document-level sentiment
  • Topic modeling
  • Aspect-based sentiment
  • Keyword extraction
  • Trend analysis
  • All four insights

Aspects measured

productpricesupportease of usereliability

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Two outputs the other use cases do not have: themes split by promoter/passive/detractor, and the sentiment read from the text compared against the actual score. If the prediction sits below the actual, there is a problem the scores are not reflecting. That is the risk numeric dashboards miss.

Worth knowing

If no score column is found the use case still runs, but the comparison section is not produced.

Open-ended survey analysis

Group hundreds of free-text answers into themes without reading them one by one.

Data

Survey responses (CSV/Excel). A text column alone is enough.

300 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Topic modelingrequired
  • Document-level sentiment
  • Intent classification
  • Keyword extraction
  • All four insights

What the report contains

Topic modeling runs first, because in open-ended answers the real question is "what was said". Then the sentiment balance of each theme and the intent of the responses (complaint, suggestion, question, praise).

Social feedback analysis

Read brand perception and competitor comparison from conversations.

Data

Social media export (CSV). With a date column, weekly trend is added.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Emotion detection
  • Topic modeling
  • Keyword extraction
  • Comparative sentimentrequired
  • Trend analysis
  • All four insights

What the report contains

Emotion detection runs here as well: separating anger from disappointment in social text is more informative than a single "negative" label. Comparative analysis puts the mentioned brands side by side.

Worth knowing

You need to enter at least two brand names for the comparison. Only you know which brands to compare, so the wizard asks.

Mobile game feedback

Player complaints speak a different language from product reviews.

Data

Google Play and App Store reviews can be pulled directly, or CSV.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Aspect-based sentiment
  • Topic modeling
  • Keyword extraction
  • Emotion detection
  • Trend analysis
  • All four insights

Aspects measured

ad frequencymonetisation balancedifficultybugs and crashesgameplaygraphics

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Ad frequency and monetisation balance are measured as separate aspects. They are the two most common complaint themes in games. Crashes are a separate aspect too, so technical problems separate from design complaints.

Restaurant & hotel

Tell whether the guest complaint is about the kitchen or the service.

Data

Google Maps or Tripadvisor export (CSV/Excel). With a score column an NPS comparison is added.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Aspect-based sentiment
  • Topic modeling
  • Keyword extraction
  • NPS/CSAT prediction
  • Trend analysis
  • All four insights

Aspects measured

tasteservice speedstaff attitudecleanlinesspriceambiencelocation

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Seven separate aspects are measured; a "bad experience" is not reduced to one score. Because staff attitude and service speed are measured separately, you can see whether the fix is training or staffing.

Worth knowing

The Google Maps connector is not yet enabled in the app; for now it works by uploading an export file.

Banking & fintech

Separate a technical fault from a fee complaint and a trust concern.

Data

Google Play and App Store reviews directly, or CSV.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Aspect-based sentiment
  • Intent classification
  • Topic modeling
  • Keyword extraction
  • Trend analysis
  • All four insights

Aspects measured

app stabilitytransaction feessecuritylending processcustomer servicecard operations

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Intent classification also runs here: in financial reviews complaints, requests and questions blur together; separating them makes the support workload visible. Security is tracked as its own aspect.

Chat & call centre

Surface recurring issues from conversation transcripts.

Data

Conversation transcripts, one FULL conversation per row.

50 / 150 comments analysed per run (default / maximum)

Steps that run

  • Intent classificationrequired
  • Document-level sentiment
  • Topic modeling
  • Aspect-based sentiment
  • 3 insights

Aspects measured

wait timeagent attituderesolutionfirst-contact resolution

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Intent classification runs first: knowing why the conversation started comes before knowing how it ended. First-contact resolution is measured as its own aspect.

Worth knowing

This use case expects one full conversation per row. If your data has one row per message (the usual shape of Zendesk/Intercom exports) the wizard warns and does not continue. In that data each message counts as an independent document, so the result describes individual utterances rather than conversations. Because transcripts are long, the row cap is deliberately low.

Employee feedback

See workplace trends from internal feedback.

Data

Employee survey or review export (CSV/Excel). If a categorical column such as department exists, a breakdown table is produced too.

200 / 1.000 comments analysed per run (default / maximum)

Steps that run

  • Document-level sentimentrequired
  • Emotion detection
  • Topic modeling
  • Aspect-based sentiment
  • All four insights

Aspects measured

managementpaycareer developmentworkloadteamflexibility

This list is not fixed, so you can edit it in the wizard to fit your own sector.

What the report contains

Six aspects measured separately. If the data has a department column, the sentiment distribution comes out as a department cross-tab. That calculation is entirely deterministic, never asked of the AI.

Worth knowing

The NPS step is deliberately absent here. The engine's current method was designed for customer satisfaction and eNPS bands differ (≥20 good, ≥50 excellent). Labelling that output "eNPS" would be wrong; it is excluded until an employee-framed method is added.

Try it with your own data

One CSV file and a few minutes is all it takes. No credit card required.

Start freeGet in touch