# Practitioner knowledge check

Ten questions on contributing and maintaining metadata at scale. You need 8 correct to pass. Unlimited retakes, with an explanation for every answer.

Question 1

1

Why is an automation preferred over manual tagging for classifying sensitive columns?

Manual tagging is not supported in CollateA rule applies to assets that do not exist yet, so classification keeps pace with ingestionAutomated tags are more accurate in every caseManual tags cannot drive access policy

Question 2

2

You are about to run a description automation across the estate for the first time. What is the safest way to start?

Run it on everything, then review the output afterwardsStart from a template rather than a blank automationDisable the schedule and trigger it manually every timeAdd a condition narrowing it to one tier or one owner, review the result, then widen

Question 3

3

What distinguishes a data contract from good documentation?

A contract is enforceable, because its terms map onto checks that runA contract applies only to external data sharingA contract is written by the consumer rather than the producerA contract is longer and more detailed

Question 4

4

You are drawing the boundary for a new data product. Which grouping is the better test?

The assets owned by the same teamThe assets that came from the same source systemThe assets that answer the same consumer questionThe assets with the same refresh schedule

Question 5

5

Reverse metadata is configured to write descriptions back to the warehouse. What new problem does this introduce?

Ingestion will stop runningThe platform becomes read onlyLineage will no longer be collectedConflict, because the same field can now be edited on both sides

Question 6

6

What does a series of profiles give you that a single profile does not?

Column level lineageThe ability to see drift, such as a null rate moving from two percent to fortyA more accurate row countAutomatic test creation

Question 7

7

When would you write a custom SQL metric instead of using a built in test?

When the rule depends on the meaning of your data, such as a settlement date never preceding a trade dateWhen the table is very largeWhen the asset has no ownerCustom metrics should be avoided wherever possible

Question 8

8

An alert has been firing several times a day for a month and the team has muted the channel. What is the actual problem?

The team is not taking quality seriouslyCustom metrics are unreliableThe threshold was never tuned after the metric went liveThe alert is routed to too few people

Question 9

9

Which split between automation and human judgment keeps a governance workflow trustworthy?

Automate detection, routing, reminders and record keeping, and leave the decision to a personAutomate everything, then audit quarterlyAutomate the decision, let humans handle exceptionsAutomate nothing, workflows should only send notifications

Question 10

10

Which case is most often left undesigned when building an approval workflow?

ApprovalNo response at allResubmissionRejection

Check my answers0 of 10 answered