Before the policies
Purview’s built-in classifiers are mostly regex on a handful of English patterns. DLP, retention and Copilot all inherit whatever they get wrong.
In a public benchmark of 89,521 files, passports were misclassified 99.8% of the time and 98.1% were unlabelled. Purview cannot classify mail at rest, which is where most privacy data lives. The bars show the shape of that gap, not live numbers from any tenant. Your own numbers come from the scan.
Per category and per workload: found, classified, labelled, missed. Filterable by directory group, with the people behind the files.
What current DLP, retention and Copilot settings protect, and the files they cannot see: mail at rest, unsupported types, on-prem shares.
A label taxonomy built from the document classes you actually have, including what should be deleted rather than labelled. We agree it with you over two workshops, and your Microsoft partner or internal team implements it.
We don’t implement Purview policies, advise on licensing or cover endpoint DLP. If you add the pilot, we write the labels back into Purview so you can watch DLP and retention act on them.
Inventory in week one. Classification five to seven days after connect. Design and report at three weeks. A label pilot adds one to two weeks.
Book a Purview baseline