Back to resources
Product note

Quallix Audit: what 'every call, scored' actually means

4 min read · 3 March 2026

A short note on what Quallix Audit produces, what it deliberately does not do, and how a score is defended when someone disagrees with it.

The claim, precisely

Every completed call ingested from your telephony is transcribed, diarised and scored against your rubric. Not a stratified sample, not calls above a duration threshold — every call the source system reports, reconciled against it.

Coverage is reported as a number you can audit, because a coverage claim that cannot be checked is a marketing claim.

What a score is made of

A score is a criterion, a result, the evidence and a confidence value. The evidence is a quoted utterance with a timestamp, so any score opens to the seconds of audio behind it.

Criteria come from your existing QA sheet plus sector conduct defaults. Changing a criterion is a configuration change, and rescoring history against a revised rubric is a normal operation rather than a project.

What Audit does not do

Audit does not talk to your customers — that is Quallix Assist. It does not assert findings it is not confident about; low-confidence segments are routed to a reviewer. And it does not replace the QA team, which remains the owner of the rubric and the arbiter of disputes.

No customer data is used to train models.