​

Tag Feedback & Approval

When auto-classification incorrectly tags data, users can report issues through a feedback system. Approved feedback automatically adds exceptions to recognizers and removes incorrect tags, creating a continuous improvement loop for classification accuracy.

​

Overview

The feedback workflow enables:

• Users to report false positives and misclassifications
• Reviewers to approve or reject feedback through task workflows
• Automatic refinement of recognizer behavior based on approved feedback
• Transparency about why tags were applied and how to correct them

​

Submitting Feedback

​

When to Submit Feedback

Submit feedback when you encounter:

• False Positive: Tag applied to non-sensitive data
• Incorrect Classification: Data is sensitive but wrong tag applied
• Overly Broad: Recognizer matches too many non-sensitive cases
• Context Specific: Valid in general but not for this specific use case

​

How to Submit Feedback

1. Hover over the auto-applied tag on a column
2. Tag details popup appears showing:
   • Tag name with “Automated” badge
   • Recognizer name that applied it
   • Confidence score
   • Timestamp
3. Click “Report” button
4. Fill out feedback form:
   • Feedback Type (required): Select the issue type
   • User Reason (optional): Choose from preset reasons or “Other”
   • Comments (optional): Provide additional context
   • Suggested Tag (optional): Select correct tag if misclassified
   
5. Submit

​

Feedback Types

​

False Positive

The tag was applied to data that is not actually sensitive. Example: An email-like pattern support@example.com appears in documentation examples, but it’s not a real email address requiring protection. What happens: If approved, the entity is added to the recognizer’s exception list and the tag is removed.

​

Incorrect Classification

The data is sensitive, but the wrong tag was applied. Example: A column is tagged as General.Email but actually contains phone numbers and should be tagged as General.PhoneNumber. What happens: If approved, the incorrect tag is removed. You can manually apply the correct tag or suggest it in the feedback form.

​

Overly Broad

The recognizer pattern matches valid cases but also produces too many false positives. Example: A pattern designed to detect SSNs also matches internal employee IDs that follow a similar format. What happens: Use this to signal that the recognizer needs refinement. Reviewers can adjust the pattern or threshold after reviewing multiple reports.

​

Context Specific

The recognizer is generally correct, but this specific case is an exception. Example: Test data in a non-production environment that looks like real PII but doesn’t require protection. What happens: If approved, this specific entity is excluded without changing the recognizer’s behavior elsewhere.

​

User Reasons

When submitting feedback, you can select from these preset reasons:

• Not Sensitive Data: Data is publicly available or not sensitive
• Wrong Data Type: Recognizer detected wrong data type
• Internal Identifier: Internal ID that looks like PII but isn’t
• Public Information: Data is from public sources
• Test Data: Synthetic or test data, not real
• Encrypted Data: Data is already encrypted/hashed
• Other: Custom reason (add comments to explain)

These reasons help reviewers understand the context and make informed approval decisions.

​

Approval Workflow

After you submit feedback:

1. Feedback status set to PENDING
2. Approval task created for tag reviewers/owners
3. Auto-approval: If you are already a reviewer for this tag, feedback is auto-approved immediately
4. Reviewers notified via their task queue
5. Reviewers assess feedback and approve or reject

​

For Reviewers: Accessing Feedback

If you are a reviewer or owner of a classification or tag:

1. Navigate to Activity > Tasks
2. Filter by task type: Recognizer Feedback Approval
3. View pending feedback submissions

​

For Reviewers: Feedback Task Details

Each task displays:

• Recognizer Name: Which recognizer applied the tag
• Feedback Type: False Positive, Incorrect Classification, etc.
• User Reason: Why the user reported it
• Comments: User’s explanation
• Submitted By: Who reported the issue
• Submitted On: Timestamp
• Entity Link: Link to the data entity

​

For Reviewers: Approving Feedback

1. Review the feedback details
2. Optional: Navigate to the entity to verify the issue
3. Click Approve button
4. System automatically:
   • Adds entity to recognizer’s exception list
   • Removes the auto-applied tag from the entity
   • Marks feedback as APPLIED

​

For Reviewers: Rejecting Feedback

1. Review the feedback details
2. Determine the feedback is not valid
3. Optional: Add rejection comment
4. Click Reject button
5. Feedback marked as REJECTED (no changes to recognizer or tags)

Use rejection when:

• The tag was correctly applied
• The user misunderstood the data sensitivity
• The classification is required for compliance

​

What Happens After Approval

When feedback is approved:

1. Entity added to exception list: The recognizer will not analyze this entity in future runs
2. Tag removed: The auto-applied tag is removed from the entity immediately
3. Feedback status updated: Status changes to APPLIED
4. User notified: The submitter can see the feedback was applied

​

What Happens After Rejection

When feedback is rejected:

1. No changes made: Recognizer and tags remain unchanged
2. Feedback status updated: Status changes to REJECTED
3. Tag remains: The auto-applied tag stays on the entity
4. Record kept: Feedback is preserved for audit purposes

​

Best Practices

​

For Users

1. Be specific in comments: Explain why the tag is incorrect with concrete examples
2. Suggest correct tags: If it’s an incorrect classification, suggest the right tag
3. Report consistently: Don’t ignore false positives - report them to improve accuracy
4. Provide context: Explain domain-specific reasons reviewers might not know

​

For Reviewers

1. Review regularly: Check pending feedback tasks weekly
2. Investigate thoroughly: Navigate to entities to verify issues before approving
3. Trust domain expertise: Users often have knowledge about data sensitivity you don’t
4. Look for patterns: If many similar reports come in, consider adjusting the recognizer
5. Add comments: When rejecting, explain why to help users understand

​

For Administrators

1. Monitor exception lists: Periodically audit exception lists for validity
2. Review auto-approvals: Check auto-approved feedback (when submitter is also reviewer)
3. Adjust recognizers: Use feedback patterns to refine recognizer configurations
4. Limit reviewers: Only grant reviewer permissions to trusted users
5. Track metrics: Monitor feedback volume to identify problematic recognizers

​

Troubleshooting

​

Cannot Submit Feedback

Issue: “Report” button not available or grayed out Possible causes:

• Tag was manually applied (not auto-applied) - feedback only works for automated tags
• You don’t have permission to submit feedback
• Tag is from a glossary term (not a classification tag)

​

Feedback Not Being Reviewed

Check:

• Tag has assigned reviewers or owners
• Reviewers have proper permissions
• Task is appearing in reviewer’s task queue

​

Tag Not Removed After Approval

Check:

• Tag was applied as “Generated” (auto-applied), not “Assigned” (manual)
• Feedback status shows APPLIED (not still PENDING)
• Refresh the page to clear cache
• Check entity history to see when tag was removed

​

Many False Positives from One Recognizer

Solution:

• Don’t just report them all individually
• Report a few examples, then notify an administrator
• Administrator should adjust the recognizer’s confidence threshold or pattern
• Use the Custom recognizers guide to refine the configuration

​

Managing Exceptions

​

View Exceptions

As a reviewer or administrator:

1. Navigate to Settings > Tags & Classification
2. Select the classification and tag
3. Go to recognizers tab
4. Click the Exceptions count on any recognizer
5. View list of excluded entities

​

Remove Exceptions

If an exception is no longer valid (e.g., test data is now production data):

1. View exceptions list
2. Click Delete on the exception
3. Confirm removal
4. Entity will be analyzed by the recognizer in future runs

​

Next Steps

Custom Recognizers

Learn how to create and configure recognizers to improve detection accuracy

Auto-Classification Workflow

Understand the complete auto-classification process