# Check knowledge quality

Read the Quality tab of Knowledge Management, from the quality score and source health to alerts, search performance and knowledge gaps, and act on each.

The **Quality** tab of Knowledge Management tells you how your knowledge sources hold up under real use: whether searches find what people ask for, where the gaps are, and which sources are unhealthy. Everything on it derives from search activity and sync health, so it fills in as people and agents start searching.

**Who can use it:** the **Admin** role or the **Knowledge Admin** grant, in a tenant licensed for knowledge. Open the Control Room, select **Knowledge**, then **Quality**.

## How it works

### Time range and refresh

The chips **24h**, **7d** and **30d** set the window for the search metrics and the trending queries; **7d** is the default, and the chosen range appears in those blocks' titles. **Updated** says when the page last fetched its data, and **Refresh** (the circular arrows) fetches it again. The page also refreshes itself every minute while its browser tab is visible.

![The Quality tab: the range chips 24h, 7d and 30d with Updated and Refresh on the right, the Quality Score gauge with its verdict beside Knowledge Health counts for Healthy, Degraded and Unhealthy, then Search Performance with Total Searches, Hit Rate, Avg Latency, Zero-Result, Positive and Negative, and Knowledge Gaps listing low-scoring questions with their best scores.](https://docs.katonic.ai/img/knowledge/knowledge-bases/quality/knowledge-quality.png)
*The Quality tab. The gauge gives the Quality Score and its verdict, Knowledge Health counts sources by health, Search Performance covers the selected range, and Knowledge Gaps lists the questions whose best match fell below the threshold.*

### Quality Score and Knowledge Health

The gauge shows the **Quality Score** from 0 to 100, averaged across your sources, with a verdict: **Good** above 80, **Fair** above 50, otherwise **Poor**. **N/A** means nothing has been scored yet. **Knowledge Health** counts your sources as **Healthy**, **Degraded** or **Unhealthy**, and adds **Unknown** for sources that have never synced. When more than a tenth of your documents may be out of date, a line under the counts gives the percentage.

### Active Alerts

**Active Alerts (N)** appears while the knowledge engine has open alerts. Each shows its message, any detail, how long ago it was raised and a severity chip; alerts from more than one source are grouped under the source. **Dismiss alert** (the cross) clears one and **Dismiss all** clears them together, confirmed by **Alert dismissed** or **All alerts dismissed**; if some could not be dismissed, the message says how many were. The block disappears when no alert remains.

### Search Performance

**Search Performance** covers the selected range: **Total Searches**, **Hit Rate**, **Avg Latency**, **Zero-Result**, **Positive** and **Negative**. Hit Rate and Zero-Result are two views of one figure and add up to 100 percent: a zero-result search returned nothing. Positive and Negative count the **Mark as relevant** and **Mark as not relevant** ratings given in Search & Test. Before any search has been recorded the block says **No search metrics available yet** and offers **Run a test search**.

### Knowledge Gaps

![The lower half of the Quality tab: Search Performance tiles and the Knowledge Gaps list on the first row, then Trending Queries ranking questions with a hit-rate bar and a count each, beside Source Health with the columns Source, Status, Docs and Last Sync.](https://docs.katonic.ai/img/knowledge/knowledge-bases/quality/knowledge-quality-gaps.png)
*Knowledge Gaps lists questions whose best match scored below the threshold, with the best score reached; Trending Queries ranks the most searched questions with a hit-rate bar and count; Source Health gives each source's Status, Docs and Last Sync.*

**Knowledge Gaps** lists the questions whose best match scored below the relevance threshold, up to eight, each with the best score it reached. A gap means people asked and the index held nothing close enough. To close one, add a source that covers the topic or extend an existing one, then reproduce the question in Search & Test. When there are none the block reads **No knowledge gaps detected.**

### Trending Queries

**Trending Queries** ranks the most searched questions in the range, up to eight, with how many times each was asked and, where the engine reports it, a hit-rate bar. A popular question with a low hit rate deserves a source. Until searches exist it says **No trending queries yet.**

### Source Health

**Source Health** lists up to ten sources with **Status**, **Docs** and **Last Sync**. The status comes from each source's health check: **Healthy**, **Degraded** or **Unhealthy**. A source that has synced before but has no health reading shows as **Healthy**; one that never synced shows its own status or **Unknown**. For a degraded or unhealthy source, select **View all N sources** to return to the **Sources** tab, open the source and read its **Runs** and **Logs** tabs. Without any source the block offers **Connect Source**.

### Next steps

The strip at the bottom leads on: **Connect a source covering a gap →** opens the Sources tab, **Adjust retrieval in Settings →** opens Settings, and **Reproduce a query in Playground →** opens Search & Test. Playground is this tab's other name for Search & Test.

## Limits and behaviour

- Before any search has happened, the whole tab is replaced by **Quality data starts when users start searching**, with **Open Playground** and **Connect a source**. Metrics need real searches, from people or from agents.
- Stale content is acted on elsewhere. When the embedding model has changed, the **Sources** tab shows a banner with **Re-embed all**; a single source is re-embedded with **Reindex**; and in a source's **Documents** tab, the refresh control on a row re-indexes that one document (**Re-index this document**, or **Retry indexing this document** for a failed one).
- Source Health shows the first ten sources; the Knowledge Health counts cover all of them.
- The score, health counts, alerts and gaps do not depend on the selected range; only Search Performance and Trending Queries do.

## Related

- [Search and test your knowledge](/knowledge/knowledge-bases/search-and-test)
- [Connect sources](/knowledge/knowledge-bases/connect-sources)
- [Knowledge](/knowledge/overview)

Source: https://docs.katonic.ai/knowledge/knowledge-bases/quality
