30-second answer

evaluate gives you a report to read. assess gives somebody else a document they can check without you.

PowerShell
m365-governance evaluate --evidence ./evidence/ --format markdown
PowerShell
m365-governance assess --evidence ./evidence/ --out ./assessment.json

Read the second line of the summary first

Text
3 rules evaluated. **1 produced an answer.**

| Outcome | Count |
|---|---|
| Fail | 1 |
| Unknown | 2 |
| Pass | 0 |

2 rules could not be decided. That is not compliance: missing evidence is a
fact about collection, not about the resource.

One failure out of three is not a third of a problem. One answer out of three is the number that describes the run, and the sentence under the table is printed by the engine rather than written by a person.

Six outcomes, and three of them are not about your tenant

pass, fail and not-applicable are claims about the resource. unknown is about the collection, invalid-evidence is about the collector, and error is about the engine.

A tool that reported unknown as pass would produce its cleanest reports from its weakest collections. This one will not, and explain prints the whole distinction from the executable.

Severity is not urgency

Every finding carries a severity the rule's author argued for in writing, and that is not the order to read them in. Attention is computed from what a finding rests on: a failure against a limit Microsoft publishes outranks a departure from advice Microsoft gives, because one breaks and the other is a decision an organisation may have taken on purpose.

The report leads with attention for that reason.

When it refuses to assemble one

assess will not write a document whose manifest would claim something the evidence does not support:

Text
refusing to assemble an assessment: <what did not line up>

A refusal, not a finding. The alternative is a signed document making a claim nobody can back.

Where to go deeper