Evaluations

An evaluation is the result of Evalua analyzing a call against your knowledge documents. Each evaluated call receives a status, detailed feedback, and a summary.

Evaluation Statuses

Evalua uses a richer set of statuses than a simple pass/fail. They fall into three groups:

AI verdicts

Passed — the call met all or most criteria. Generally no action needed.
Failed — the call did not meet quality standards. Review with the agent for coaching.
To Review — the AI was uncertain. A Supervisor needs to make the final call.

Human-review verdicts

When a Supervisor reviews a "To Review" call, the status moves to one of:

Review Pass — the supervisor judged the call as passing
Review Fail — the supervisor judged the call as failing

See Reviewing Call Results for the workflow.

Non-evaluated states

Unevaluated — the call hasn't been evaluated yet (auto-evaluate is off, or it was reset)
Missing Recording — the audio couldn't be retrieved from the phone system
No Knowledge — the workspace has no knowledge documents, so there's nothing to evaluate against
Limit Exceeded — the company/user is over their transcription quota for the period

A status of Processing is used internally while a call is being transcribed and evaluated; you don't typically see it as a final state.

For AI-graded calls, status is set by the workspace's quality thresholds — score ranges that map a percentage of criteria met to Passed, To Review, or Failed. Supervisors configure thresholds in Settings → General.

Example: 80-100% met = Passed, 50-79% = To Review, below 50% = Failed.

What an Evaluation Contains

Overall status — one of the values above
Criterion-by-criterion analysis — whether each Feedback criterion was met
Overview data — structured data extracted by Overview documents (agent name, customer name, etc.)
Specific feedback — explanations of why criteria were met or missed
Summary — a brief overview of the call (visible to agents)
Full transcription — the complete call text with speaker labels (supervisors and admins only)

Automatic vs Manual Evaluation

When Auto Evaluate is enabled in workspace settings, calls are evaluated automatically after transcription completes. This is the default workflow.

Supervisors can also manually re-evaluate calls — useful when criteria or Context Documents changed after the call was first graded. Re-evaluation takes 2-5 minutes.

Human Review Workflow

Calls in To Review status need supervisor input:

Open the call from History
Read the transcription and the AI's analysis
Mark the call Review Pass or Review Fail

This keeps a human in the loop on the cases the AI couldn't decide confidently, without requiring manual review of every call.