A proof of concept is being turned into a governed production service. Which monitoring capability should be added that the proof of concept did not need?
Select an answer to reveal the explanation.
Short Explanation
A proof of concept only has to work once, in a demo. A production service has to keep working, which means you need ongoing tracking of data and model drift so you know when live behaviour has diverged from the validated model.
Full Explanation
The correct answer is the option about ongoing tracking of data and model drift, so the team knows when live behaviour has diverged from the validated model. A proof of concept is evaluated once and retired; a production service runs continuously against live, changing data, so it needs ongoing drift monitoring that the proof of concept never required.
The option about a single accuracy score captured once, to record in the project closure report, is incorrect because a single captured score is a point-in-time record, not the ongoing signal production needs.
The option about a code-quality linter run during the build pipeline is incorrect because a linter checks code style, not model behaviour, and would already be useful even at the proof-of-concept stage.
The option about a user satisfaction survey distributed after each release is incorrect because a satisfaction survey measures user perception, not the underlying data or model change that causes performance to shift.
For the exam, connect the move from prototype to production with the need for continuous, not one-off, behaviour monitoring.