Setting up Scorecards

A Scorecard ties several Detectors together into a scoring system.

To configure, go to Scoring setup → Scorecards tab.

A Scorecard defines which Detectors to run, how to evaluate the result for each one, how many points each output is worth, and what minimum score a call needs to pass. It also controls which calls get scored — by assigning the Scorecard to specific campaigns or verticals. When a call from a specific campaign is processed, every Detector in the Scorecard runs.

The results are combined into a:

• QC Result (Pass or Fail)

• QC Score (percentage)

Before creating a Scorecard, make sure you've created the Detectors you want to include. See Setting up Detectors.

Managing your Scorecards

Using templates

Kaliper provides a library of pre-built Scorecards for common use cases. Toggle Show templates ON to browse them.

Templates are read-only. To use one, open it and click Add to Campaign — this creates an editable copy in your org.

Limits

• Maximum 500 Detectors per organization

• Maximum 50 Detectors per Scorecard

• Maximum 50 Scorecards per organization

Browsing and filtering

The table has a Quick search field for instant name filtering.

The collapsible filter panel (labeled "Search" in the UI) lets you narrow by verticals, campaigns, specific detector, and whether templates are shown.

Table columns:

Manually creating a Scorecard

Fields

Click Add Scorecard at the top of the Scorecards tab.

Name and Description A display name is required. Description is optional.

Verticals & Campaigns Controls which calls this Scorecard applies to. Assign it to specific campaigns, to whole verticals (which covers all campaigns in that vertical), or a mix of both.

If you select a campaign whose vertical is already selected, the campaign is removed from the direct list automatically — the vertical assignment already covers it.

If a Scorecard was not assigned to any campaigns or verticals — it never runs on any calls.

Adding Detectors

Click the Add Detectors field to open the picker. Browse by type, category, or existing Scorecard. Each Detector you add gets a configuration row.

Adding a Detector from the Detector editor applies default values (score 10, not Critical, not Conversion Related). Always review the configuration row in the Scorecard.

Configuring each Detector

Critical and Conversion Related flags

Both flags are per-Scorecard — the same Detector can be Critical on one Scorecard and ordinary on another.

Critical

A Critical Detector is a hard gate. If the output doesn't match the target, the entire call fails — regardless of how many other Detectors passed. The QC Score still gets calculated, but QC Result is forced to Fail.

Before marking a Detector Critical, we recommend to run it for a week as a regular weighted Detector. Review the calls where the output didn't match the target. If the AI was wrong on more than a few percent, refine the prompt before promoting to Critical.

Conversion Related

A Conversion Related Detector tells Kaliper that this Detector's output is the conversion signal for the call. When the output matches the Target Answer, Conversion Detected on the call is set to Yes — overriding the AI's default conversion verdict from transcript analysis. See Scoring and Conversion detection for the full priority logic.

If multiple Conversion Related Detectors fire on the same call — any of them is enough for the call to be counted as converted.

Conversion Related does not affect the QC Score, the Pass/Fail verdict, or any other scoring output.

Disposition Detectors with multiple targets

For Disposition Detectors you can select multiple items as targets — each gets its own row with its own score value. This lets you award different point values to different outcomes (e.g., "Converted" worth more than "Callback Requested").

Threshold

The Threshold (%) sets the minimum score a call must reach to pass.

• 0% — all calls pass unless a Critical Detector fails

• 80% — calls must earn at least 80% of possible points to pass

• 100% — every scored Detector's output must match its Target Answer to pass

A few starting points: begin at 70–80% if you're not sure what's realistic, run the Scorecard for a week, then adjust based on your actual pass rate. Use Critical Detectors for hard requirements rather than raising the threshold to 100% — it's a more precise tool. A threshold of 0% with only Critical Detectors is valid when you care only about hard gates, not a weighted score.

Scorecard-level context

The Advanced section contains a single field: Context prompt.

This is background information sent to the AI alongside every Detector prompt — the campaign type, what the agent is required to do, compliance rules, or domain-specific terminology.