> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orca-klavest.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Analytics

> Long-term trends across every job in your workspace — quality over time, recurring issues, GDPR exposure, file leaderboards, and AI cost tracking.

## Overview

The Analytics page is the workspace-wide view of your data quality posture over time. Where [job comparison](/comparison) answers "what changed between two specific runs," analytics answers "what's the trend across the last 30 days, 90 days, or all time."

Open **Analytics** from the sidebar. Everything is org-scoped — workers see analytics filtered to jobs they have access to, admins see the full org view.

## What it shows

Analytics is organised into five panels, each backed by its own endpoint group.

### Quality

The quality panel tracks how clean your data is over time:

| Chart                 | What it shows                                                              |
| --------------------- | -------------------------------------------------------------------------- |
| **Quality over time** | Daily/weekly aggregate quality score across all jobs                       |
| **Quality delta**     | Period-over-period change with significance markers                        |
| **Null rates**        | Average null percentage per period, broken down by column type             |
| **Recurring issues**  | Issues that have appeared on multiple files — your "long tail" of problems |
| **Top issues**        | The biggest single issues by row count or severity                         |

The recurring issues view is the most actionable: it surfaces the patterns worth fixing at the source rather than per-file.

### Files

The files panel shows volume and per-file behavior:

| Chart                | What it shows                                            |
| -------------------- | -------------------------------------------------------- |
| **File volume**      | Files processed per day/week, by source type             |
| **File leaderboard** | Worst-quality files in the period                        |
| **File history**     | Score timeline for a single named file across re-uploads |
| **Filenames**        | All filenames seen, with last-scan timestamps            |

The leaderboard is the fastest way to find "the file that always breaks contracts" so you can prioritize root-cause work.

### GDPR

The GDPR panel tracks your PII exposure:

| Chart                | What it shows                                 |
| -------------------- | --------------------------------------------- |
| **GDPR exposure**    | Count of PII columns detected per period      |
| **GDPR file detail** | Per-file breakdown of detected PII categories |

Use this for compliance reviews and to verify that GDPR scanning is actually catching what it should.

### Insights

The insights panel breaks down quality issues by category:

| Chart                     | What it shows                                                     |
| ------------------------- | ----------------------------------------------------------------- |
| **Issue breakdown**       | Distribution of issue types (nulls, duplicates, format, outliers) |
| **Category distribution** | What semantic categories your data is composed of                 |
| **Column heatmap**        | Which columns have the most issues across all files               |
| **Column detail**         | Drill into a single column's issue history                        |
| **Recent activity**       | Latest scans, completions, and contract violations                |

The column heatmap is excellent for "we always have problems with the customer table" investigations.

### Value & AI cost

The value panel quantifies what ORCA is actually doing for you:

| Chart                | What it shows                                                                      |
| -------------------- | ---------------------------------------------------------------------------------- |
| **Value summary**    | Estimated time and cost saved by ORCA's automated checks                           |
| **Value benchmarks** | Your usage vs typical workspaces of similar size                                   |
| **AI costs**         | Token consumption per period, broken down by operation type                        |
| **AI health**        | Classification confidence trends — is the AI getting better or worse on your data? |
| **AI drift**         | Whether the AI's classification decisions are stable over time                     |

Use AI costs to forecast monthly token spend and AI drift to catch the rare case where the AI's behavior on your data shifts (usually because your data shifted).

## Filtering

Every panel supports the same filters at the top of the page:

* **Date range** — last 7 days, 30 days, 90 days, or custom
* **Source** — filter to a single connected data source
* **File pattern** — match by filename glob
* **Job status** — filter to complete only, or include partial/failed

Filters affect every panel simultaneously so you can scope the entire view.

## API access

Analytics is a thin layer over many endpoint groups. The exact shapes are documented in [API endpoints](/api-reference/endpoints), but the family is:

```bash theme={null}
GET /api/v1/analytics/quality-over-time
GET /api/v1/analytics/quality-delta
GET /api/v1/analytics/null-rates
GET /api/v1/analytics/recurring-issues
GET /api/v1/analytics/top-issues

GET /api/v1/analytics/file-volume
GET /api/v1/analytics/file-leaderboard
GET /api/v1/analytics/file-history
GET /api/v1/analytics/filenames

GET /api/v1/analytics/gdpr-exposure
GET /api/v1/analytics/gdpr-file-detail

GET /api/v1/analytics/issue-breakdown
GET /api/v1/analytics/category-distribution
GET /api/v1/analytics/column-detail
GET /api/v1/analytics/column-heatmap
GET /api/v1/analytics/recent-activity

GET /api/v1/analytics/value-summary
GET /api/v1/analytics/value-benchmarks
GET /api/v1/analytics/ai-costs
GET /api/v1/analytics/ai-health
GET /api/v1/analytics/ai-drift
```

All endpoints accept the same date range and filter parameters as the UI. All return empty data on error rather than failing — analytics are non-critical and never block the rest of the app.

## Tips

* **Start with the quality-over-time chart.** A flat line is healthy. A jagged line means upstream variability worth investigating. A steady decline means you have a regression problem.
* **Use recurring issues as your backlog.** Anything that shows up on three or more files is worth fixing at the source, not per-file.
* **Watch AI drift after big upstream changes.** When your source schema changes, drift spikes are normal — but they should settle within a few scans. Sustained drift means the AI's confidence is genuinely lower on the new data.
* **Set up a weekly digest email.** Combine analytics with [scheduled reports](/reports) so the highlights land in your inbox without anyone opening the dashboard.
* **Treat AI cost as a budget signal, not a performance metric.** Cost going up means you're scanning more data, not that ORCA is getting more expensive per scan.

## What's next?

* [Job comparison](/comparison) — when you need a focused two-job diff instead of trends
* [Reports](/reports) — schedule analytics summaries for your team
* [Contracts](/contracts) — convert recurring issues into enforced rules so they stop showing up in analytics