A governed production model needs a decision on how often its thresholds should be reassessed after launch. What is the best practice?
Select an answer to reveal the explanation.
Short Explanation
The model's context shifts - new users, new data patterns - and a threshold set once at launch stops matching reality unless you revisit it. You need to reassess thresholds periodically and whenever the model's context or usage pattern changes.
Full Explanation
Operational risk shifts as usage and context change, so thresholds need periodic and event-driven reassessment. The best practice is to reassess thresholds periodically and whenever the model's context or usage pattern changes, so they continue to reflect real operational risk. Setting thresholds once at launch and never revisiting them is incorrect because a static threshold drifts out of alignment with the model's actual risk over time. Reassessing only if the model is completely retired and replaced is incorrect because waiting until retirement means the threshold is never useful during the model's active life. Reassessing on a fixed five-year cycle regardless of usage changes is incorrect because a fixed long cycle ignores meaningful changes that happen well before five years pass. For the exam, treat threshold reassessment as ongoing, tied to both schedule and context change.