# How To Compare Quality Trends Across Agents And Categories

## Objective

Compare QA patterns across a consistent time period and evaluation scope, then inspect the underlying work before acting.

## Before you begin

You need company-scope QA reporting access: **Quality: Manage** or **View Only**. Select a time period with enough completed evaluations for the groups you intend to compare.

## Resolution

1. **Open the report.** Go to [Reporting · QA · Automated QA](https://wfm.zenops.cx/service/reporting/qa-analysis/automated-qa). **QA · General** shows run volume and processing health; **Automated QA** shows evaluation quality.
2. **Hold the scope steady.** Choose the date range, category, and run origin. Use the same filters for each comparison. The default origin includes automatic, requested, test, and batch work; narrow it when comparing operational performance. Retries replace the attempt they repeat rather than counting as a second evaluation.
3. **Compare the trend.** Review the quality trend, score distribution, low-score rate, and critical-fail rate. Check the evaluated sample behind an average before treating a small change as meaningful.
4. **Compare categories and criteria.** Use **Category performance** and **Criteria performance** to find differences in work types and standards. Select a criterion to inspect its trend. Categories may use different published scorecards, so a raw score difference does not necessarily mean the teams faced the same rubric.
5. **Compare agents where available.** Use **Agent performance** and its trend for agent-level evaluations. The report keeps agents with too few evaluations visible but does not rank them. Agent batch runs grade one selected agent and have no overall ticket grade, so ticket-level metrics are unavailable when only that origin is selected.
6. **Inspect the underlying runs.** Follow a metric or category link into [QA Grades](https://wfm.zenops.cx/service/quality). Read individual evidence and manager decisions before choosing a coaching action. For an agent conversation, use that agent's current published result and [Coaching](https://wfm.zenops.cx/service/quality/coaching).

The report is based on AI assessments and can include setup tests or historical batches depending on the chosen origin. A published result is the version the agent actually sees. See [How To Find Coaching Opportunities In QA Results](/help/articles/qa-how-to-find-coaching-opportunities-in-qa-results/) for the follow-up workflow.

> Canonical page: https://zenops.cx/help/articles/qa-how-to-compare-quality-trends-across-agents-and-categories/
> Audience: For managers and analysts
> Summary: Compare QA trends across a consistent period, origin, agents, categories, and criteria.