EDM Classifier in Microsoft Purview: No Match Detected After Successful Indexing

Hello everyone,

I’m working on an Exact Data Match (EDM) classifier in Microsoft Purview using the new experience, and I’m facing an issue where the classifier does not return any matches during testing, even though all steps were completed successfully.

What I Did

1. Created EDM Classifier:
   • Primary element: CPF (ndodocumentodeidentificacao), with ignored delimiters (.,-).
   • Text fields configured as case-insensitive.
   • Phone field configured to ignore spaces and hyphens.
2. Detection Rules:
   • High Confidence: CPF + Name + Email + Phone.
   • Medium Confidence: CPF + Name OR Email.
   • Proximity: default (300 characters), Wide Mode.
3. Downloaded schema, ran ValidateData → Passed.
4. Performed hash and upload using EdmUploadAgent.exe → Completed successfully.
5. Status in Purview: Index Complete.

The Problem

When I test using EDM Classifier Test with a .docx file containing exactly one row from the original CSV (CPF, Name, Email, Phone, etc.), the results are:

• Base SIT (Brazil CPF) detects the CPF.
• EDM SIT → No match detected.

What I’ve Verified

• Document saved as a real .docx (not just renamed).
• Data matches the CSV exactly (same formatting, same delimiters).
• Indexing status shows Complete.

Questions

• Has anyone experienced this issue?
• Could this be related to tokenization, proximity, or confidence rules?
• Are there any additional configurations or limitations I should be aware of?

Environment

• Microsoft Purview (new EDM experience).
• EDMUploadAgent installed on Windows Server.
• CSV file: UTF-8, comma-delimited, header matches schema.

Any insights or troubleshooting tips would be greatly appreciated!