Question 950
A gold metrics pipeline reports wrong totals even though all Databricks Jobs are green. Engineers have applied Compare against fixed baseline fixtures and revert unsafe transformation. Which result establishes the required behavior: An optimization change altered numerical results unexpectedly?
Explanation
Correct Answer
The acceptance criterion for “An optimization change altered numerical results unexpectedly” is Regression tests confirm unchanged business measures after repair; this is the expected evidence after Compare against fixed baseline fixtures and revert unsafe transformation.
Incorrect Answers
The transform runs and exact revenue sample results match pertains to a different control (Restore validated numeric typing and strict cast handling) and a different need (A nullable amount column was accidentally converted to string and subsequent SUM failed). It cannot establish that Compare against fixed baseline fixtures and revert unsafe transformation met the stated goal.
Gold latest event date matches the expected processing date would demonstrate the separate action Add the missing job dependency and a freshness assertion, which addresses “Data from yesterday appears in current dashboard due to missing upstream refresh dependency.” It does not verify the requirement “An optimization change altered numerical results unexpectedly” following Compare against fixed baseline fixtures and revert unsafe transformation.
That observation is a valid indicator for “Source rows are duplicated by retries and the gold SUM is too high” when using Deduplicate on business event ID before aggregation and repair affected periods. The scenario instead tests “An optimization change altered numerical results unexpectedly”; the relevant result is Regression tests confirm unchanged business measures after repair.