> For the complete documentation index, see [llms.txt](https://docs.wonderchat.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.wonderchat.io/setup-guides/understanding-resolution-rate-and-satisfaction.md).

# Resolution Rate & Satisfaction (CSAT)

Your analytics dashboard shows two headline numbers for every chatbot: **Resolution Rate** and **Satisfaction (CSAT)**. Both are measured **per conversation**, and each conversation is scored in one of two ways:

1. **Customer feedback** — a 👍 or 👎 left on the chat. This always takes priority.
2. **AI evaluation** — if there's no feedback, an AI judge scores the conversation automatically, about **15 minutes after it goes quiet**.

The AI judge reads the whole conversation and looks at things like the customer's tone (were they satisfied or frustrated?), whether the chatbot was able to answer what was asked, and whether the chat reached a helpful conclusion.

***

## Resolution Rate

Of the conversations where it was clear whether the customer was helped, the percentage that were **resolved**.

```
Resolution Rate = Resolved ÷ (Resolved + Unresolved)
```

Conversations that aren't a real question — greetings, spam, or nonsense — are marked **unclear** and left out, so they don't distort your number.

{% hint style="info" %}
A conversation is also marked resolved when a customer leaves a 👍, when a live agent resolves the chat, or when an automated workflow completes successfully.
{% endhint %}

***

## Satisfaction (CSAT)

The **average satisfaction score** across all scored conversations in the selected date range, on a scale of 1 (extremely dissatisfied) to 5 (extremely satisfied).

```
Average CSAT = sum of all scores ÷ number of scored conversations
```

The score comes from the customer's 👍/👎 where given, otherwise from the AI judge. Conversations with no score yet are excluded, and if nothing has been scored in the selected window the dashboard shows **N/A**.

***

## Good to know

* **Recent chats may not show yet.** Scores are assigned about 15 minutes after a conversation goes quiet, so very recent or still-active chats won't be counted until then.
* **Feedback beats the AI.** If a customer leaves 👍 or 👎, that's the source of truth for both metrics. The AI judge only fills in conversations with no feedback.
* **Resolution and Satisfaction are different.** One asks *"did we solve it?"*, the other *"did they feel good about it?"* — and they can diverge. A chat can be resolved but rated low, or unresolved but rated high.
* **The date range and chatbot selector apply.** Both metrics reflect the conversations created within the date range and chatbot(s) selected on your dashboard.

***

If you have any more questions, feel free to reach out to us at <mark style="color:purple;">**<support@wonderchat.io>**</mark>