Feedback insights
Feedback insights aggregate every customer rating and written comment so you can track CSAT, NPS, and sentiment in one place.
Feedback insights roll up every rating and comment customers leave on a conversation. Use the page for weekly reporting, to track CSAT and NPS over time, and to read the comments that explain what's really going on. Go to Insights → Feedback to open it.
Summary cards
The top of the page shows the headline numbers for your selected date range.
| Card | What it shows |
|---|---|
| Rating distribution | A bar chart of how many ratings landed at each star (or thumbs up/down). |
| CSAT | Customer satisfaction score, calculated from positive ratings. |
| NPS | Net Promoter Score, shown when a widget asks the standard NPS question. |
| Sentiment over time | Average customer sentiment per day, derived from written comments. |
NPS only appears if at least one of your widgets is configured to ask the standard NPS question. If you don't see it, that's why.
Verbatim comments
Below the cards is a list of every written comment customers left in the date range. Each entry shows the rating, the comment text, the widget it came from, and a link to the conversation.
The list is the most useful part of the page. Numbers tell you something is off; comments tell you what.
Filter to low ratings only
Tick the low-rating filter to focus on chats that didn't go well.
Skim the comments
Look for repeated words or topics — that's where to focus your next round of document work.
Open a chat
Click the conversation link next to a comment to read the full transcript. From there, follow Working with a conversation to act on it.
A few low-rated chats on the same topic are the strongest signal you can get. Two thumbs-down on "shipping to Canada" outweighs a hundred neutral chats.
Filters
The filters at the top of the page apply to both the summary cards and the verbatim list.
Filter by widget
See feedback for a single widget when you want to know whether a recent change moved the needle.
Filter by handoff
Split feedback between AI-only chats and chats that escalated to a human. The two often score very differently.
Filter by rating
Quickly switch between "lowest first" for triage and "highest first" for testimonials.
Export to CSV
Click Export in the top right to download the full verbatim list as a CSV. The export includes the rating, comment, conversation ID, widget, timestamp, and the customer's identifier when one is on file. Use it to pipe feedback into a reporting tool or share it with a stakeholder.
The export respects your active filters and date range, so you can pull a clean monthly slice without editing the file.
Quick wins workflow
Most teams turn weekly feedback review into one or two concrete improvements.
Set range to last 7 days
Use the Last 7 days preset.
Filter to lowest ratings
Anything 1–2 stars (or thumbs down).
Group by topic in your head
Skim the comments and tally the top two themes.
Ship one fix
Write or update a document, tighten a persona, or attach a skill. One change per week beats a backlog.
What's next
Document insights
Document insights surface the articles your AI cites most and the ones it ignores, so you can find knowledge gaps and prune dead weight.
Cost insights
Cost insights break down what you're spending on AI by feature, widget, and project, with day-over-day deltas so you can catch spikes early.