> ## Documentation Index
> Fetch the complete documentation index at: https://docs.argide.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Satisfaction (CSAT)
> Ask end users to rate closed conversations and track the results
Satisfaction surveys ask end users to rate a conversation once it closes. Ratings for the AI agent and for your support team are kept apart, so you can see how each one is doing.
## Turn on surveys
Open **Settings** → **Satisfaction**.
Turn on the toggle under **Rating survey**. The label changes to **Rating survey is on**. Surveys are off by default.
**Ask after an inactivity close** is on by default. Turn it off to survey only conversations that ended with a clear resolution or a support agent's close.
Edit **Question** and click **Save**. It appears above the rating faces in the widget. The default is "How was your experience?", and it can be up to 200 characters.
Toggles save as soon as you change them. Only an organization admin can change these settings.
## When end users are asked
A survey goes out when a conversation closes:
| How it closed | Who gets rated |
| - | - |
| The end user confirmed the AI agent's answer helped | The AI agent |
| The conversation closed for inactivity, with **Ask after an inactivity close** on | The AI agent |
| A teammate clicked **Close** after replying in the conversation | Your support team |
A few cases send nothing:
* A teammate closes a conversation without having replied in it.
* Neither the agent nor a teammate ever replied.
* The end user's last message is more than 7 days old.
Once a teammate replies, the AI agent is no longer surveyed in that conversation. A conversation can still end up with two ratings: one for the AI agent before the handover, and one for your team afterwards.
If a conversation is reopened and closes again, the end user sees the same survey again rather than a new one. This stops once they've left a comment or 14 days have passed since they were first asked.
See [Inbox](/conversations/inbox#open-snoozed-and-closed) for how conversations close.
## What end users see
The widget shows a rating card at the end of the conversation with your question and five faces:
| Face | Rating | Counts as |
| - | - | - |
| 🤩 | **Amazing** (5) | Positive |
| 😃 | **Great** (4) | Positive |
| 😐 | **Ok** (3) | Neutral |
| 🙁 | **Bad** (2) | Negative |
| 😠 | **Terrible** (1) | Negative |
Picking a face submits the rating straight away. A **Tell us more…** field then lets them add a comment of up to 500 characters.
The end user can change their rating until they leave a comment, or for 14 days after they were asked. After that the card reads "This conversation can no longer be rated".
In the inbox, the rating shows on the conversation's row, as a marker in the **Chat** tab, and under **Rating** in the **Details** tab.
## The Satisfaction report
Open **Conversations** → **Satisfaction**. Pick **Last 7 days**, **Last 30 days**, or **Last 90 days**. The default is 30 days.
Every metric except **CSAT request rate** counts surveys by when they were **asked**, not when they were answered. A survey asked on the last day of a window and answered the next day still counts in that window.
### Scores
| Metric | Formula |
| - | - |
| **Overall CSAT score** | Positive answers (Great or Amazing) ÷ answered surveys |
| **Overall DSAT score** | Neutral and negative answers (Ok, Bad, or Terrible) ÷ answered surveys. Always 100% minus the CSAT score. |
| **AI agent CSAT score** | Positive answers ÷ answered surveys, for the AI agent's surveys |
| **Support agent CSAT score** | Positive answers ÷ answered surveys, for your support team's surveys |
Each tile also shows the counts behind it, such as "42 out of 50". A score reads **—** when no survey was answered.
**CSAT score over time** charts the overall score by **Day** or **Week**. A day or week with no answers leaves a gap in the line rather than dropping to 0%.
### Ratings and remarks
* **Positive remarks**, **Neutral remarks**, and **Negative remarks** count the answers in each group.
* **Conversation ratings - by rating** shows how many answers each face received.
* **Conversation ratings** lists every answered survey, newest first: the rating, the comment, the end user's name, who handled the conversation, and when they answered. Filter by face, and click the time to open the conversation.
### Survey coverage
| Metric | Formula |
| - | - |
| **CSAT request rate** | Conversations started in the window that were asked for a rating at any point ÷ all conversations started in the window |
| **CSAT response rate** | Answered surveys ÷ surveys sent in the window |
| **Surveys sent** | Surveys asked in the window, split between the AI agent and your support team |
| **Surveys answered** | Surveys sent in the window that the end user answered, split the same way |
A conversation rated for both the AI agent and your team counts as two surveys.
### Comparison with the previous period
Each tile compares itself with the previous period of the same length, just before the current one. For example, with **Last 30 days** selected, it compares with the 30 days before that.
* Scores and rates change in percentage points, such as "▲ 4 pts vs previous 30 days".
* Counts change by the number of answers or surveys.
* Green means the change is good and red means it's bad. Lower is better for **Overall DSAT score** and **Negative remarks**. **Neutral remarks**, **Surveys sent**, and **Surveys answered** aren't colored.
A tile reads "No data in the previous 30 days", with the number of days you selected, when there's nothing to compare against.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.