Home › Per Kohavi 'Trustworthy Online Controlled Experi…
Per Kohavi 'Trustworthy Online Controlled Experiments', SRM (sample-ratio mismatch) means which?
AObserved bucket split deviates from designed split
BCPU sampling rate differs from the GPU sampling rate
CAudio sample rate differs from the video frame rate
DNetwork packets are reordered in flight in transit
Answer & Solution
Correct answer: A. Observed bucket split deviates from designed split
Kohavi: SRM is an early-warning that experiment plumbing is broken — randomisation bias, telemetry loss skewed per arm. χ² test flags it.
Related questions
Per Kohavi + Riot, the CUPED variance-reduction technique for A/B tests uses which?Per Riot + Activision postmortems, an EVENT-SCHEMA REGISTRY (e.g. Iteratively, Segment ProPer Amplitude + GameAnalytics docs, the recommended way to spec EVENT names is which?Per Microsoft + Booking experimentation platforms, NOVELTY EFFECT in A/B testing refers toPer Riot Games engineering blog, GUARDRAIL METRICS in live experiments are which?Per Mixpanel + Amplitude docs, an IDENTITY-RESOLUTION step in analytics joins which?Per GameAnalytics + Tableau best practices, the SAMPLING approach for high-volume telemetrPer Unity Analytics + GameAnalytics, a COHORT in retention analysis is typically defined b