Skip to content

Quality Assurance

Quality Assurance (QA) in MiaRec is the systematic evaluation of conversations against your organization's quality standards. The primary method is Auto QA — AI-powered scoring of conversations — with manual evaluation available as a secondary method. This page defines the QA concepts; the User Guide QA chapter covers the day-to-day workflow.

Scorecard

A scorecard is the template that defines what "good" looks like for a conversation. It is organized into sections (for example Greeting, Problem Solving, Closing), each containing questions such as "Did the agent introduce themselves properly?". Questions can be answered with choices (for example Yes / No / N/A), or with a numeric score.

Questions and sections carry weights, so the items that matter most count the most, and the scorecard has a passing threshold — the minimum score a conversation must reach to pass.

For Auto QA, each question is written in plain language, optionally with extra guidance that clarifies for the AI what counts as a correct answer — for example, explaining that a "proper" introduction includes thanking the caller and offering help.

Where you see it: scorecards are managed under the QA menu. See Scorecard designer.

Auto QA

Auto QA uses AI to score conversations against a scorecard automatically. The AI reads the transcript, answers every question, and provides a justification for each answer with evidence from the conversation — including where in the call the evidence occurs — so reviewers can verify the result rather than take it on faith.

Because scoring is automatic, Auto QA is not limited to a small sample: it can evaluate every conversation that matches the criteria your organization sets, giving complete coverage of agent performance and customer experience.

Where you see it: auto-scored evaluations appear in the Evaluations list and on the conversation's QA tab, like any other evaluation. See Understanding Auto QA in the User Guide and Enabling Auto QA in the Administration Guide.

Manual evaluation

A manual evaluation is a human review: a supervisor picks a conversation and a scorecard, listens to the call, and answers the questions. Manual evaluation is included with the Quality Assurance product and is useful for calibration, sensitive cases, or scorecards you prefer to keep human-scored. An evaluation can also be assigned to another evaluator with a due date.

See Manually evaluating a conversation.

Scoring at a glance

Whether an evaluation is automatic or manual, the score is calculated the same way:

  • Each answered question earns points; question and section weights determine how much each contributes.
  • Questions answered N/A are excluded from the calculation entirely — they neither help nor hurt.
  • The result is a percentage score for each section and for the whole evaluation.
  • The evaluation passes if the overall score meets the scorecard's passing threshold. Individual answer choices can be configured as failing: selecting one forfeits the points for that question and the questions that follow it in the section or form.

The exact calculation, with worked examples, is in Score calculation.

Review and override

Auto QA is designed for human oversight. A supervisor reviewing an auto-scored evaluation can:

  • Read the AI's answer and justification for each question.
  • Override any answer they disagree with — the score recalculates, and the original AI answer is preserved alongside the override for transparency.
  • Leave feedback on the AI's answer; scorecard owners use this feedback to refine question wording and guidance so future scoring improves.
  • Add section-level feedback and overall comments: what went well, what could be improved, and any additional feedback.

See Reviewing evaluations.

Coaching

Evaluations turn into agent improvement through feedback:

  • The written feedback on an evaluation (strengths, improvement areas, per-section comments) gives the agent concrete, evidence-backed guidance.
  • Notes can be attached to a conversation — including anchored to specific moments of the transcript — so coaching comments point at exactly what was said.
  • Agents can view their own evaluations, so feedback reaches them directly in the product.