A scorecard is the structured rubric Ender Turing uses to evaluate a conversation. It groups your quality criteria into categories, and each category contains one or more points (also called scoring parameters) that an evaluator — a human reviewer, or the AutoQA engine — answers when scoring a conversation.

The Scorecards screen is where you build and maintain those rubrics for your workspace:

Scorecards live separately from teams. Linking a scorecard to a team and choosing each team's default scorecard happens from the Users & Agents area, not from the Scorecards screen itself.

Open Settings from the main navigation, then click Scorecards.

The page opens on the list of scorecards in your workspace. A top-right toggle switches between Active and Archived views.

Only users whose role includes the team-management permission see the Scorecards entry in the Settings menu and can open the list. Users without that permission do not see the menu item and cannot reach the page.

Within the page, access splits into two levels:

A user with only the scorecard-management permission can reach a scorecard's edit page from a direct link and inspect its structure, but saving any change still requires the team-management permission — the Save call will be rejected otherwise. The standard Settings → Scorecards navigation entry only appears for users who have the team-management permission.

A scorecard that you cannot edit because it is managed by an external Template appears with a read-only indicator and shows a "view" icon instead of an edit pencil. You can still open it to inspect its structure.

Each row shows:

A quick filter at the top of the page lets you narrow the list by:

Each row's action area offers:

From the list, click + Create card in the top bar. A dialog appears with two tabs:

After creating, you land on the scorecard's edit page.

Click + Import card in the top bar. Choose an .xlsx file and click Upload. The platform parses the file and creates a scorecard with the categories and points described in the spreadsheet. If the file is in the wrong format, or a scorecard with the same name already exists, an inline error explains the problem and no scorecard is created.

After a successful import, Ender Turing opens the new scorecard so you can fine-tune it.

The edit page has two columns.

The left column contains the scorecard's categories, each shown as an expandable panel:

The right column is the scorecard's settings sidebar:

The top bar of the edit page contains:

When you save changes that would alter historical scoring results — adding a new point, removing a point, changing a point's max score, changing its partial-score values, flipping its critical flag, or changing the N/A buttons behaviour — Ender Turing asks you to confirm before proceeding. The confirmation explains that historical scoring results may change significantly. Other edits (renaming a point, updating descriptions, reordering, changing partial-scoring eligibility without changing values) save without that warning.

If you try to leave the page with unsaved changes, you are asked to confirm that it is safe to discard them.

When you save a scorecard, Ender Turing checks the points you are removing against any Enders (automations) that reference them. If at least one Ender — active or archived — uses one of the points you are removing, the save is blocked and the platform lists which points must stay. This is stricter than archiving the whole scorecard, because an archived Ender can be re-enabled later and would still expect those point IDs to exist.

While editing, attempts to delete points whose IDs are referenced elsewhere open a Unable to remove current point(s) dialog that lists the dependent entities (such as Enders) point by point, so you can remove or unwire those references first.

Toggling Set as Default on a scorecard makes it the workspace-wide default scorecard that reviewers see first when they open the scoring panel. Only one scorecard can be the default at a time; turning a different scorecard into the default replaces the previous one.

Note that a workspace default is separate from the per-team default scorecard that teams can have configured in the Users & Agents area. The per-team setting takes priority when a team is set, while the workspace default acts as a fallback.

Toggling Use as Automated lets AutoQA score conversations against this scorecard. After at least five AI/LLM auto-scored conversations have been human-validated, the Scorecard Accuracy indicator shows the weighted accuracy and gives actionable guidance text (for example, "The AI/LLM instructions need improvement" or "Validated accuracy is reasonably high").

On an automated scorecard, each point row also shows its own accuracy value. It is measured per instruction version and counts only the human-validated AutoQA scores produced by the version currently in use. Until that version has five of them, the row shows Not enough data instead of a percentage, with a tooltip explaining that at least 5 validations are needed. Hovering the value gives the same kind of guidance text as the scorecard-level indicator.

Click the value to open the Accuracy details dialog, which holds the accuracy chart, the point's definition versions, and the tools for improving what AutoQA executes. That workflow is covered in its own article: Scorecard Accuracy Improvement.

If the point has neither observations nor an automated scorecard behind it, no accuracy element is shown at all.

Each definition version stores an Instruction the AI uses — the compact text AutoQA actually executes. Editing a point's description and saving the scorecard creates a new version built from that text and puts it straight into use, which also resets the point's accuracy to Not enough data. Renaming a point updates its label without creating a new version.

The name and description in the scorecard form always remain your text — activating an AI-improved version changes only the instruction AutoQA executes, never what you typed. To try new wording without changing what AutoQA does today, use Edit definition inside Accuracy details instead, which saves an inactive version. See Scorecard Accuracy Improvement.

The list's Archive action on each row moves a scorecard out of the active list. Archived scorecards still exist (their historical scores are preserved) but they are hidden from scoring pickers and from the active scorecard list.

Switch between Active and Archived in the top-right toggle to view either set.

Deleting a scorecard is permanent and is intentionally restricted:

If teams or dependent entities still reference the scorecard, the delete icon is disabled with a tooltip listing what needs to be removed first (for example, "team", "conversations", "Enders"). When all dependencies are clear, click the Delete icon and confirm in the dialog to remove the scorecard.

On the edit page's top bar, the Excel-export icon downloads the scorecard structure — categories, points, score values, critical flags, and descriptions — as an .xlsx file. You can edit that file and re-import it as a new scorecard if you want to clone a scorecard between workspaces or share the rubric for offline review.

Copying an automated scorecard inside the workspace carries its current point text and active runtime instructions, but not its reviewed examples, accuracy observations, candidate evaluations, or version history. Spreadsheet export/import carries the point text and creates a fresh simple definition whose rule and runtime instruction use that text. Tuning evidence is specific to the workspace conversations on which it was collected.

Each scorecard in the list has a Labels cell where you can attach coloured labels. Click the + to add an existing label or create a new one. Once attached, you can filter the list by one or more labels using the quick filter at the top of the page.

Automated scoring depends heavily on how unambiguous the scorecard is: