Assign A/B variants to experiment rows before analysis run

Your experiment rows arrive without variant labels, causing split analysis to stall and noisy metrics. Ensure labeled experiment records so analysts can run clean A/B reports before the next analysis run.