Scorecard Templates

A scorecard template is a reusable set of evaluation criteria. Templates shape how meeting summaries are structured and how AI interviews are scored.

What a scorecard template is#

A template is a named collection of criteria. Each criterion has:

• A name — e.g., "Technical depth", "Communication", "Ownership".
• A description — What "good" looks like for this role; how the AI should interpret responses.
• A rating scale — Typically a 1–5 or 1–4 numeric score with labeled anchors.

When a template is attached to a meeting, the AI evaluates the transcript against each criterion and writes a short justification plus a numeric score.

Creating a template#

Go to Templates → Scorecards (or the equivalent location in Settings, depending on your workspace).

1. Click New template.
2. Give it a name and, optionally, a description of the role type it's for.
3. Add 3–5 criteria. Fewer is better than more — broad templates score less reliably.
4. For each criterion, write a description that tells the AI what to look for. Be concrete.
5. Pick a rating scale (most teams use 1–5).
6. Save.

Editing and versioning#

Templates are versioned. Editing a template creates a new version; meetings that were already scored with the previous version keep their original scores and context.

This matters because it means you can tune a template over time without retroactively changing past evaluations. If you want a past meeting re-evaluated with a new version, regenerate its summary from the meeting detail page (see Transcripts and Summaries).

Attaching a template to a stage#

Templates aren't applied directly to individual meetings — they're attached at the pipeline-stage level, and all meetings produced from that stage inherit the template.

From the pipeline editor (see Pipeline and Stages):

1. Click Edit on a stage.
2. Under Scorecard template, pick the template you want.
3. Save.

From that point, new AI interviews or meetings produced in that stage will be scored against the template.

Reviewing and dismissing scores#

On any meeting detail page, the scorecard section shows the AI's score and justification for each criterion.

You can:

• Accept — Leave the AI's evaluation as-is; it becomes part of the candidate's record.
• Override — Enter your own score and justification. Your override is preserved separately from the AI's original output.
• Dismiss — Mark the AI's evaluation as invalid (e.g., the call was cut short and the AI had insufficient data). Dismissed scores don't affect candidate summaries or downstream triage.

Best practices#

• Keep templates tight. Three to five criteria score more reliably than ten.
• Write concrete descriptions. "Communicates clearly under ambiguity" beats "Good communication".
• Reuse across jobs. A well-tuned "Senior engineer" template can serve every senior engineer role you run.
• Review early scores manually. The first few candidates through a new template will tell you whether the criteria are producing useful differentiation. Tweak if not.