# Know what was evaluated, why it scored, and what needs to improve.

> Quality applies configured, versioned scorecards to eligible ingested interactions. It preserves criterion-level rationale, findings, coverage, and the improvement work that follows.

## Key Specifications
- **Target Audience:** Quality leaders, support operations, and risk teams
- **Primary Inquiry:** Can we move beyond samples without hiding the denominator?

## Core Architecture & Mechanism
### 1. Define
Configure the scorecard version, criteria, source scope, and eligible population.

### 2. Queue
Resolved or imported interactions enter the audit pipeline when they meet the eligibility rules.

### 3. Explain
Store criterion scores, rationale, evidence, findings, severity, and audit state.

### 4. Improve
Route findings into review, coaching, alerts, and repeated-pattern analysis.

## Frequently Asked Questions
### Q: Can we use our own scorecard?
A: Yes. Scorecards and criteria are configured for the entitled tenant scope and preserved by version.

### Q: Can historical work be evaluated?
A: Controlled imports and Quality ingestion contracts support selected chat, email, and audio sources in supported formats.

### Q: Does full coverage mean every interaction?
A: No. It means coverage of the population that was configured, ingested, and eligible for the applicable scorecard.

## Claim Boundaries
- OpenOS does not claim to audit every interaction everywhere; coverage is bounded by configured eligibility and ingestion.
- Sentiment is one signal and is not represented as customer satisfaction.
- Automated scoring does not eliminate calibration, review, or policy ownership.
- Outcome improvement must be measured after an intervention; a finding alone is not a verified business outcome.

---
Canonical Reference: https://www.openost.com/audit-agents