Krisp AI Speech Analytics: QA Scorecards

QA Scorecards allow you to evaluate calls against a structured set of questions that map directly to your team's QA rubric. After each call, Krisp automatically scores the transcript against your scorecard and provides a justification for each answer.

Unlike call metrics, which use an AI-generated 1–3 scale to evaluate agent behaviors, QA Scorecards are built from your own questions with Yes, No, or N/A answer types and custom point weights, mirroring how your QA team would manually review a call.

How QA Scorecards work

A QA Scorecard is made up of one or more question groups. Each group contains a set of questions, and all questions within a group are evaluated together. After the call ends, the transcript is analyzed against the scorecard and a final score is calculated automatically.

Question groups

Questions are organized into groups, such as Standard Greeting, Verification, or Issue Resolution. Groups can be used to mirror your existing QA form structure.

Answer types and scoring

Each question is answered as Yes, No, or N/A. Point values are configurable per answer type. For example, a Yes answer might carry positive points, a No might carry zero or negative points, and N/A might carry no weight.

Auto-zero

Critical questions or groups can be flagged with an auto-zero setting. If a question or group with auto-zero enabled is answered No, the total score for the call automatically becomes zero, regardless of other answers. This is typically used for questions that represent absolute requirements.

Total score

The final score for each call is calculated based on the points earned across all questions. A QA possible score is also calculated, representing the maximum score achievable given the answers provided (accounting for N/A responses).

Score justifications

For each question, Krisp provides a justification explaining why the answer was Yes, No, or N/A based on the call transcript. This allows supervisors to review the reasoning behind each score without listening to the full recording.

Setup and customization

There is no default QA Scorecard. Each team's scorecard is defined separately, based on their QA requirements. Scorecards support flexible configurations to match different QA programs, including weighted questions, group-level auto-zero for critical failures, and N/A handling for questions that do not apply to certain call types.

QA Scorecard configuration is handled by the Krisp team and is not available for self-serve setup. For setup instructions and initial configuration, see Getting started with Krisp AI Speech Analytics as an Admin.