# AI feedback analysis

Every answer a customer writes in a flow is classified automatically by sentiment, category and intent, so you can read the themes instead of every comment.

## How each piece of feedback is classified

| Label | Values | Answers |
|---|---|---|
| Sentiment | Positive · Neutral · Negative | How the customer feels |
| Category | General · Pricing · Functionality · Usability · Support · Refund · Content · Junk | What the comment is about (Junk filters empty or meaningless answers) |
| Intent | Opinion · Issue · Request · Suggestion | What the customer wants you to do with it |

Classification runs as sessions arrive. Both the follow-up answers from the survey and free-form feedback are included.

## The feedback feed

Each card shows the customer, the date, the cancellation reason they picked, the follow-up question and their answer, the free-form feedback, and the three labels.

![A feedback card: reason, feedback, and the sentiment / category / intent labels.](https://churnsolution.com/wp-content/themes/astra-child/assets/images/docs/fb-card.png)

You can filter the feed by time period, sentiment, intent, user plan, flow and subscription status, search for a user or a word in the feedback, and export the filtered list.

## Breakdowns

- **Categories:** share of feedback per category (for example Pricing 12 %, Functionality 11 %).
- **Intent:** share per intent (opinion, issue, request).
- **Sentiment by category:** for each category, the positive / neutral / negative split, so you see that pricing comments are mostly negative while usability comments are mostly neutral.
- **Sentiment over time** is on the main [Analytics](/docs/cancellation-flows/analytics/) dashboard.

## Custom categories

Categories can be tailored to your product, and the model can be pretrained on your own historical feedback on request. Contact the team.