Retraining Trainable Classifiers in Microsoft Purview

Answer Correct answer: A — Only unpublished classifiers (Classifier1) can be retrained; published classifiers must be removed and recreated.

You have a Microsoft 365 E5 subscription that has the trainable classifiers shown in the following table. Which trainable classifiers can you retrain? -

Classifier1 only Correct Answer
Agreements and Classifier1 only
Classifier1 and Classifier2 only
Agreements, Finance, Classifier1, and Classifier2

microsoft purview trainable classifiers retrain published classifier sc-400 custom classifier retraining rules

Community Votes

71%

29%

71% of anonymous learners picked answer A. Votes are pick records left by other test-takers — they are not the verified answer.

Community Insight

Tests the distinction between unpublished (editable/retrainable) and published (immutable) custom classifiers. The trap is assuming published classifiers can be updated directly.

Determines which trainable classifiers can be retrained based on their publishing status in Microsoft Purview Information Protection. The page clarifies that only unpublished classifiers can be retrained, while published ones must be recreated.

Choosing Option D or C because users assume all 'trainable' classifiers are retrainable regardless of state, ignoring the 'Published' flag which locks the model.

Community Discussion (6 comments)

Domza 👍 7

Note: Pre-trained classifiers cannot be re-trained

JimboJones99 👍 1 Selected: A

A - reasoning is same as NICKTON81 and techpam

NICKTON81 👍 4 Selected: A

A is correct: Retraining classifiers Retraining published custom classifiers is no longer supported. If you need to improve the accuracy of a trainable classifier you've published, remove the classifier and start over with larger sample sets. To improve the accuracy of an unpublished classifier, review the test results, update the data set with additional data, and restart the training. https://learn.microsoft.com/en-us/purview/trainable-classifiers-learn-about#retraining-classifiers

techpam 👍 4

Retraining published custom classifiers is no longer supported. If you need to improve the accuracy of a trainable classifier you've published, remove the classifier and start over with larger sample sets. https://learn.microsoft.com/en-us/purview/trainable-classifiers-learn-about#retraining-classifiers

emartiy 👍 2 Selected: C

Based on I understand from below link, a Custom Trainable Classifier does not need be used in any policy, only publishing is enough to retrain it. So C is seems correct. https://learn.microsoft.com/en-us/purview/trainable-classifiers-learn-about#:~:text=When%20you%20publish%20the%20classifier%2C%20it%20sorts%20through%20items%20in%20locations%20like%20SharePoint%20Online%2C%20Exchange%2C%20and%20OneDrive%2C%20and%20classifies%20the%20content.%20After%20you%20publish%20the%20classifier%2C%20you%20can%20continue%20to%20train%20it%20using%20a%20feedback%20process%20that%20is%20similar%20to%20the%20initial%20training%20process.

Futfuyfyjfj 👍 3

I agree based on the answer options, however besides the option I would say only classifier 2 can be retrained: https://learn.microsoft.com/en-us/purview/classifier-how-to-retrain-content-explorer ‘A classifier must already be published and in use before it can be retrained.’

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Expert Analysis

Why the Answer Is Correct

The correct answer is A (Classifier1 only). According to Microsoft's official documentation for Purview Information Protection, retraining published custom classifiers is no longer supported. To improve accuracy, you must remove the classifier and create a new one with larger sample sets. Therefore, Classifier2 and Agreements, being published, cannot be retrained.

Why the Other Options Are Wrong

Options B, C, and D include published classifiers (Classifier2 and Agreements). Since published models are locked to ensure consistency across policies, they do not support the retraining workflow. Only unpublished classifiers allow for dataset updates and restart training.

Community Comment Notes

Community members like Domza and NICKTON81 correctly noted that pre-trained/published classifiers cannot be retrained. Some learners (like emartiy) argued for C, suggesting unpublished status is enough, but missed that Classifier2 is explicitly listed as Published. The consensus aligns with the vendor's update that retraining is restricted to the unpublished phase.

Official Reference

https://learn.microsoft.com/en-us/purview/trainable-classifiers-learn-about#retraining-classifiers

Exam Strategy

Always check the 'Status' column in trainable classifier tables. If it says 'Published', the answer is usually 'No' for modification or retraining questions. Remember: Unpublished = Edit/Retrain; Published = Remove & Recreate.

Frequently Asked Questions

Why can't I retrain Classifier2?

Classifier2 is published. Microsoft Purview does not support retraining published custom classifiers to maintain policy stability.

How do I fix an inaccurate published classifier?

You must remove the existing classifier and create a new one with additional training samples to improve accuracy.

Related Analysis

← Back to SC-400 Study Guide