Evidence Reports
Evidence Reports are periodic risk-scoring snapshots that evaluate every agent × data resource pair in your workspace. Each pair is scored on six weighted signals and ranked from highest to lowest risk, giving you a prioritized view of where to focus governance attention. Reports can be reviewed, verified or rejected, and exported as CSV or PDF.
Generating a Report
- Go to Agent Trust → Evidence Reports.
- Pick a Period start and Period end date for the reporting window.
- Click Generate report.
The report enters a Generating status while Bigeye computes scores. The page auto-refreshes every few seconds until generation completes. Once finished, the report moves to Draft status and is ready for review.
Report Statuses
| Status | Description |
|---|---|
| Generating | Scoring is in progress. |
| Draft | Report is ready for review. |
| Verified | A team member has reviewed and approved the report. |
| Rejected | A team member has rejected the report. |
| Failed | Scoring could not complete. Try generating a new report. |
By default, the report list hides rejected reports. Check Show rejected to include them.
Reviewing Findings
Click a report's date to open its findings page. Each row represents one agent × resource pair, ranked by total risk score (highest first). The findings table includes:
- Rank
- Agent: links to the agent's detail page
- Resource type: Column or Table
- Resource path: links to the resource's catalog page
- Total score: composite risk score out of 100
- Confidence: Full if all six signals have measured data, Partial if any signal fell back to an estimate
- Per-signal scores: individual scores for each of the six scoring signals
Scoring Signals
Each agent × resource pair is evaluated on six signals. Signals are normalized and combined into a weighted total score (0–100).
| Signal | Weight | What it measures |
|---|---|---|
| Sensitivity | 20% | Whether data sensitivity scans found positive findings for this resource. |
| Data Quality | 20% | Ratio of open data quality issues to active metrics on the resource. |
| Active Issues | 20% | Count of open issues, weighted by priority (P1 issues count more heavily than P3). |
| Access Frequency | 15% | How often the agent has accessed the resource, on a logarithmic scale. |
| Downstream Impact | 15% | Number of downstream dependencies in the resource's lineage graph. |
| Business Criticality | 10% | The resource's popularity score. |
Score Details
Click Score details on any finding row to open a breakdown of how that pair's total score was calculated. The modal shows each signal's individual score and whether it was based on Measured data or an Estimated default.
When Bigeye doesn't have measured data for a signal — for example, if sensitivity scans haven't run on a resource — it applies a neutral midpoint estimate (0.5) and marks the finding's confidence as Partial. Scores marked with an asterisk (*) in the findings table indicate at least one estimated signal.
Verifying and Rejecting Reports
Once a report is in Draft status, you can verify or reject it from the findings page using the Actions dropdown.
- Verify marks the report as reviewed and approved. Verified reports can still be rejected later.
- Reject marks the report as rejected. Rejection is final and cannot be reversed.
Exporting Reports
From the Actions dropdown on the findings page, you can export a report as:
- CSV with all finding data, including raw counts and per-signal scores.
- PDF in a formatted, landscape-oriented tabular layout.
Both formats include the full set of findings and are named with the report's date range.
Updated about 2 hours ago