Sample ratio mismatch
For a two-arm experiment, the SRM diagnostic compares observed visitor counts with the planned allocation using a chi-square approximation. It is unavailable when expected counts are too small or the design is not supported by that diagnostic.
An unexpected allocation ratio can signal targeting, identity, consent, exposure or instrumentation problems. Investigate before interpreting a treatment difference.
Measurement checks
Confirm that identifiers are stable, exposure happens after rendering, event names match goals, deduplication IDs are stable for retries, and consent is explicitly connected. Check configuration delivery and whether the conversion window has matured.
Empty or incomplete results
An empty result is not evidence of no effect. First check the selected environment and date range, then verify assignment and actual recorded exposure and goal events in your own integration.