Purview Scans are creating Schema Classifications but not Data Classifications

Jarrod Henningsen 20 Reputation points
2025-02-27T19:18:12.1333333+00:00

Full scans have been run on an ADLS Gen2 storage account containing an example csv file with PII data at both the Auto and Level3 ScanLevel. Schema Classifications were found for PII related rules such as physical address, full name, etc, but no classifications were created. Scan rule set AdlsGen2 (System) was used for the scan. Why are Data Classifications not being generated.

Microsoft Security | Microsoft Purview
{count} votes

Answer accepted by question author
  1. Anonymous
    2025-03-03T14:11:16.4333333+00:00

    Hello @Jarrod Henningsen ,

    I understand your concern regarding the data classifications.

    The below possible scenarios might be the reasons for the above behavior. Do let us know if not.

    • If the sample size is too small or not representative, the system may not confidently apply a classification. You need to ensure that your data contains sufficient and representative samples of the sensitive information you expect Purview to classify.
    • As per this MS Documentation, for system classification minimum match threshold value is set at 60% and cannot be changed. Cross-check whether your input csv data met this threshold or not.

    Coming to your custom classification, please try the below possible workaround methods while creating it.

    • There might be some hard-coding mistake while giving the Regular expression and you can consider generating the automatic Regular expression by giving a sample csv file containing all required classification values. If needed, you can test the classifying rule with another set of csv file containing sample test cases for this classification. enter image description here
    • You can try the Dictionary classification which also needs a csv file with single column of containing required values.

    I hope this information helps.

    If the answer is helpful, please click Accept Answer and kindly upvote it. If you have any further questions about this answer, please click Comment.

    1 person found this answer helpful.

1 additional answer

Sort by: Most helpful
  1. Jarrod Henningsen 20 Reputation points
    2025-03-03T23:18:32.9066667+00:00

    Thanks Rakesh. I will try testing with different custom classifications rules and datasets

    0 comments No comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.