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

# Coaching Scorecards

> Build weighted scorecard templates and review rep performance on individual conversations.

## Overview

Coaching scorecards are reusable templates that define the criteria managers use
to evaluate rep performance on a single conversation. Each scorecard becomes a
saved **review** attached to that conversation. Scorecards support multiple
question types, weighted criteria, optional sections, conditional applicability,
and per-criterion AI assist, so you can grade a discovery call, a support
ticket, or a chat with the same system.

## How it works

A scorecard is a **template** managed in Settings, then **filled** as a review on
a conversation's Coaching tab. A template holds a list of **criteria**, each with
a type, a relative **weight**, and optional settings. The five criterion types
are:

| Type          | What the reviewer does                | How it scores                                                      |
| ------------- | ------------------------------------- | ------------------------------------------------------------------ |
| Numeric (0–N) | Enters a number from 0 to a max score | The number entered, divided by the max score (default max 5)       |
| Yes / No      | Picks Yes or No                       | Yes = 100%, No = 0%                                                |
| Pass / Fail   | Picks Pass or Fail                    | Pass = 100%, Fail = 0%                                             |
| Single choice | Picks one labelled option             | The chosen option's points, divided by the highest option's points |
| Text note     | Writes a qualitative note             | Informational, excluded from scoring                               |

The rollup is a **weighted percentage**. Weights are *relative*. They normalise
at calculation time, so they do **not** need to total 100%. A criterion marked
**N/A** is dropped from the denominator entirely. A
**Pass / Fail** criterion with **Auto-fail review** enabled, when failed, pins
the whole review to 0% and shows a **Failed** badge.

Criteria can be grouped under **sections**, and you can restrict where a
scorecard appears using **"Applies when"** rules (see below). Toggle **Required**
to force the scorecard to be submitted before a matching conversation can be
completed.

### Applies when (applicability rules)

Each rule narrows the scorecard to matching conversations. All rules are AND'd;
an empty list means the scorecard applies to every conversation. Fields:

* **Source type**: meeting, phone, widget/chat, Slack, or email.
* **Disposition / outcome**: the conversation's outcome.
* **Channel**: specific channel ids.
* **Duration (seconds)**: greater than / less than a threshold.

A **Required** scorecard whose rules don't match a conversation is *not* required
for that conversation.

### AI assist and prefill

Criteria with **AI assist** on can be drafted by the model. On a conversation's
Coaching tab, **AI Prefill** reads the transcript and proposes values for those
criteria. Suggestions are never auto-submitted; the manager always reviews and
clicks **Submit Review**. Criteria with AI assist off are skipped by prefill, and
any value the model returns that doesn't fit the criterion type becomes N/A for
you to fill manually.

## Using it

<Steps>
  <Step title="Open Coaching Scorecards">
    Go to **Settings → Coaching → Coaching Scorecards** and click **New Scorecard**.
  </Step>

  <Step title="Add criteria">
    Give the scorecard a name. For each criterion, set a **Label**, pick a **Type**, and set its **Weight**. Configure
    type-specific options (max score for numeric, the option list for single choice) and toggle **Required**, **AI
    assist**, or **Auto-fail review** (Pass / Fail only) as needed.
  </Step>

  <Step title="Group and scope (optional)">
    Assign criteria to **sections**, then use **Applies when** to restrict the scorecard to certain source types,
    dispositions, channels, or call lengths. Toggle the top-level **Required** switch if reviewers must submit it.
  </Step>

  <Step title="Score a conversation">
    Open a completed conversation's **Coaching** tab, pick the scorecard, score each criterion (or use **AI Prefill**),
    add notes, and click **Submit Review**.
  </Step>

  <Step title="Confirm the result">
    The submitted review shows a weighted percentage badge (or **Failed** if an auto-fail criterion failed). Expand it
    to see the per-criterion breakdown.
  </Step>
</Steps>

<Note>
  Editing a template does not change reviews already submitted with it. Existing reviews keep the criteria they were
  scored against.
</Note>

## Troubleshooting

| Symptom                                           | Likely cause                                                            | What to check                                                                                                                    |
| ------------------------------------------------- | ----------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| Coaching tab shows a lock icon                    | The conversation hasn't finished processing                             | Conversation must be **completed**, close and process it so the transcript and analytics generate first                          |
| Scorecard not offered on a conversation           | Applicability rules don't match                                         | Review the **Applies when** rules; confirm the conversation's source type, disposition, channel, and duration satisfy every rule |
| "No scorecard templates yet" on the Coaching tab  | No active templates exist                                               | Create one in **Settings → Coaching → Coaching Scorecards**; archived templates don't appear                                     |
| Can't save: "needs at least two labelled options" | A single-choice criterion is incomplete                                 | Single-choice criteria require at least two options, each with a label                                                           |
| Review shows **Failed** at 0%                     | An auto-fail Pass / Fail criterion was marked Fail                      | Open the review and check the failing criterion; auto-fail pins the whole review to 0%                                           |
| Weights "don't add up to 100%"                    | Expected. Weights are relative                                          | Weights normalise at calculation time; they don't need to total 100%                                                             |
| AI Prefill returns blanks for some criteria       | Those criteria have **AI assist** off, or the model's value was invalid | Turn on **AI assist** for the criterion; invalid values fall back to N/A for manual entry                                        |

<Note>
  When reporting an issue, include the request id shown in the error toast. It links the in-app error to the server
  logs.
</Note>

## Related

* [Coaching Trends](/comms/coaching/trends)
* [Conversations & Threading](/comms/conversations/overview)
* [Recording](/comms/voice/recording)


## Related topics

- [Coaching Trends](/comms/coaching/trends.md)
- [Sales](/start/use/sales.md)
- [What the assistant can do](/ai/tools/library.md)
- [Change Log](/product/updates.md)
- [Voice & Calls](/comms/voice/overview.md)