An assessor’s office plans to replace a production valuation helper with a newer model. Which practice best controls that change?
Select an answer to reveal the explanation.
Short Explanation
Let the new model challenge the champ under controlled tests before it takes the field. Retraining on validation data contaminates your evaluation, and logging lineage or publishing dashboards for review won't give you controlled tests and a rollback path before cutover.
Full Explanation
Assessor valuation helpers in production affect civic outputs that residents and staff rely on, so model replacement needs controlled comparison. Comparing champion and challenger models under controlled evaluation before replacing the production valuation helper best controls that change: the incumbent remains the champion until the challenger proves superior on agreed metrics with a rollback path. That practice protects integrity and service continuity during municipal model upgrades.
Retraining the newer valuation helper on validation sales records before replacing the production valuation helper fails because it trains on data meant for unbiased evaluation, contaminating the validation set and inflating accuracy while still substituting retraining for controlled champion-challenger comparison with approval and rollback. Recording model version lineage in the inventory registry before replacing the production valuation helper fails because lineage documentation alone, even when done beforehand, provides provenance without proving the challenger is superior on agreed metrics and without establishing comparison or rollback protection. Publishing parcel valuation accuracy dashboards for stakeholder review before replacing the production valuation helper fails because transparency reporting alone does not control the change—it shares accuracy information without requiring controlled champion-challenger evaluation and approval before cutover.
Exam caveat: Champion/challenger controlled evaluation before production replacement is the lifecycle change-control pattern; retraining on validation data, lineage recording alone, and dashboards for review without staged comparison are distractors.
Operational check: Run shadowed or staged challenger evaluation against the champion on valuation test sets, require recorded metric comparison and approval before cutover, and retain a rollback path to the prior helper version.
Prefer to study offline? Printable PDF study book + test engine ($7.50, unofficial). These questions stay free here.