CG
You work on a digital product team and an A/B test has just come back with an unexpected imbalance between control and treatment traffic. The primary metric looks promising, but the allocation does not match what was planned, and there are signs of possible data quality issues in the readout.
How would you handle sample ratio mismatch or other data quality issues in the experiment readout, and how would you decide whether the result is trustworthy enough to act on?
