> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vorel.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# CSAT surveys

> Ask customers how it went: per-channel delivery, thresholds that avoid surveying the wrong people, and the report.

CSAT is the customer telling you how it went, as opposed to the platform's own opinion of its work.
Both appear in reporting, and where they disagree is usually the interesting part.

## Turning it on

Under **Settings, then CSAT**:

| Setting                       | What it controls                                                       |
| ----------------------------- | ---------------------------------------------------------------------- |
| **Master switch**             | Whether any surveys are sent                                           |
| **Channels**                  | Which of WhatsApp, web chat, email and voice get surveyed              |
| **Delay**                     | How long after the conversation closes to ask                          |
| **Link lifetime**             | How long the rating link stays valid                                   |
| **Minimum customer messages** | How much the customer has to have said before they are worth surveying |
| **Comment box**               | Whether to collect free text alongside the score                       |
| **Question wording**          | What you actually ask, in English and Arabic                           |

The minimum-message threshold is the setting people forget. Without it, a wrong number that sent two
words gets a satisfaction survey.

## How the ask is delivered

The survey rides the channel the customer already used:

| Channel  | Delivery                                                                       |
| -------- | ------------------------------------------------------------------------------ |
| WhatsApp | A message on the same thread                                                   |
| Email    | An email with a rating link                                                    |
| Web chat | A rating bar inside the widget, since a visitor has no address to message      |
| Voice    | A WhatsApp message to the caller's number, because a caller has no text thread |

## When it fires

Every way a conversation closes produces the same survey: an operator closing it, an operator
resolving it, a call ending, or the inactivity sweep. There is one close path, so there is one place
the survey is enqueued and no way for a close route to quietly skip it.

## Rating

The customer rates 1 to 5 through an unguessable single-use link, or in place in the widget. The
capture route is public and does not require the customer to have an account, and rating twice does
not create a second response.

## The report

**Quality, then CSAT** shows:

* Average score and response rate
* The full 1 to 5 distribution, which tells you far more than the average
* A per-channel breakdown
* A daily trend with the change against the previous period

A per-channel split matters because CSAT differs sharply by channel. Voice usually scores higher
than email, and comparing a blended number month to month while your channel mix shifts tells you
nothing.

## Reading it honestly

* **Response rate first.** A 4.8 average from 3% of customers is three enthusiasts, not a signal.
* **Watch the 1s and 2s, not the average.** Detractors tell you what is broken; the average hides
  them.
* **Compare with what the platform thought.** A conversation the quality scoring rated well and the
  customer rated 1 is worth reading in full.

## Events

A recorded rating emits a `csat.recorded` [webhook](/integrations/webhooks), so scores can be pushed
into your own reporting or your CRM.

## Retention

Survey responses are retained for 365 days, longer than transcripts, because a satisfaction score is
an aggregate measure rather than conversation content. See
[Data retention](/security/data-retention).