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

# Review human ratings in Feedback

> Collect thumbs up and thumbs down on agent replies and workflow runs, then triage, assign, and export them from Studio.

Feedback is the inbox for ratings people leave on Scout output. A visitor can mark a Copilot reply helpful or unhelpful. Someone on your team can rate a workflow run. Those ratings show up together under **Evaluate → Feedback**, or at [studio.scoutos.com/feedback](https://studio.scoutos.com/feedback).

Use Feedback when you want a person's judgment on a specific reply. Use [Evaluations](/agents/evaluations) when you want an automatic score on live traffic.

Until the first rating arrives, the page says **No feedback has been logged yet**.

## How ratings get here

### Copilot replies

On a [Copilot](/agents/copilot) reply, the visitor uses the thumbs up or thumbs down control under the message. Scout stores that rating on the agent session.

In the table, a thumbs rating shows as **👍 Good** or **👎 Poor**. The score is 100% for thumbs up and 0% for thumbs down.

The same session in the agent's **History** tab shows a thumbs icon: **Rated helpful by the user** or **Rated unhelpful by the user**.

### Workflow runs

Open the workflow, open **Logs**, and open a run. Use the feedback control on the run details. Rate **Response Quality** as **Excellent**, **Good**, **Okay**, or **Poor**, and add a comment if you want one.

That row links to the workflow. In the details panel, **View Workflow Run** opens the log for that run.

## Review a rating

The table lists the newest ratings first. Columns:

| Column | What it shows |
| - | - |
| **Input** | **👍 Good** or **👎 Poor** for a thumbs rating. Other ratings show the score as a percentage. |
| **Score** | The same score as a percentage, with a pass mark at 50% or above. |
| **Priority** | **Low**, **Medium**, **High**, or **Critical**. |
| **Status** | **New**, **Reviewing**, **Resolved**, or **Dismissed**. |
| **Target** | The workflow name, or **View session** for an agent conversation. |
| **Assignee** | Who on the team owns the follow-up. |
| **Created** | When the rating was submitted. |
| **Comment** | The comment, if one was included. |

Click a row to open the details panel. From there you can change **Status** and **Priority**, open the source session or workflow run, and read the user message, the agent response, and the full comment when Scout captured them. The panel also shows who submitted the rating. External visitors are marked **External**.

New ratings arrive as **New**, with **Medium** priority.

<Steps>
  <Step title="Open Feedback">
    In Scout Studio, open **Evaluate** in the left sidebar and choose **Feedback**.
  </Step>

  <Step title="Open the rating">
    Click the row. The details panel opens on the right.
  </Step>

  <Step title="Jump to the source">
    For an agent session, click **View session** or **View Session**. Studio opens that conversation on the agent. For a workflow run, click **View Workflow Run**.
  </Step>

  <Step title="Close it out">
    Set **Status** as you work it. Click **Mark Resolved** when you are done. **Reopen** moves a resolved item back to **Reviewing**.
  </Step>
</Steps>

## Filter the list

**Search feedback** matches comments on the rows currently loaded. The table loads the 50 newest ratings that match your filters, so search does not look through older comments sitting outside that set.

**Filter** narrows what Scout loads:

* **Score** — **Positive** is 50% or higher. **Negative** is below 50%. Thumbs up is positive. Thumbs down is negative.
* **Priority** — **Low**, **Medium**, **High**, or **Critical**.
* **Status** — **New**, **Reviewing**, **Resolved**, or **Dismissed**.
* **Assignee** — one workspace member, or **Unassigned**.
* **Date Range** — when the rating was created.

**Clear All** removes every filter. If filters hide every row, the table says **No feedback found**.

## Assign and delete

In the **Assignee** column, pick a workspace member or **Unassigned**.

To remove a rating, open the row menu and choose **Delete**. Confirm **Are you sure you want to delete feedback?** Deleting a rating does not delete the session or the workflow run.

## Export agent feedback

**Export** downloads a CSV of agent-session ratings that match the current filters and comment search, including rows beyond the 50 on screen. Workflow-run ratings stay in the table. They are not in the file.

The file is named `agent-feedback-YYYY-MM-DD.csv`. Columns:

`id`, `created_at`, `agent_id`, `agent_name`, `session_id`, `score`, `status`, `priority`, `assignee`, `comment`, `user_message`, `agent_response`

`score` is a number from 0 to 1. `assignee` is the member's name. `user_message` and `agent_response` are the text Scout stored with the rating, so someone outside Studio can see what was asked and what the agent said.

If nothing matches, Studio reports **No agent feedback to export** and does not download a file. If the export is too large to finish in one file, Studio tells you the file is partial.

**Export** is unavailable on the empty page, before any rating exists.

## Who can open Feedback

| Role | Submit a rating | Open **Feedback** |
| - | - | - |
| **Member** | Yes | No. **Feedback** is hidden, and a direct link is blocked. |
| **Editor**, **Admin**, **Owner** | Yes | Yes. They can filter, assign, update, delete, and export. |

If **Feedback** is missing under **Evaluate**, your role does not include read access. Ask an Admin or Owner to grant it. See [Roles](/settings/roles).

## Feedback and evaluations

[Online evaluations](/agents/evaluations) score conversations with an LLM judge or a code scorer. A human-review scorer cannot run on its own inside an online suite.

Feedback is the place those human ratings go: one person, one reply or run, then a status your team can work. Use both when you want an automatic quality signal and a queue of ratings a person actually left.

## Next steps

<CardGroup cols={2}>
  <Card title="Copilot" icon="message-bot" href="/agents/copilot">
    Embed the agent whose replies visitors rate.
  </Card>

  <Card title="Evaluations" icon="gauge-high" href="/agents/evaluations">
    Score live conversations automatically, separate from these human ratings.
  </Card>

  <Card title="Observability" icon="chart-line" href="/agents/observability">
    Open the session behind a rating and read the trace.
  </Card>

  <Card title="Roles" icon="user-shield" href="/settings/roles">
    See who can submit a rating and who can open the inbox.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.