When you consume CloudBees impression callbacks, always check impressionData.targeting first. targeting=false means the user got the flag's default value because no targeting rule matched, not because they were bucketed into a variant. If your analytics treats default-served impressions as exposures, control and treatment counts inflate and experiment reads go wrong. Filter analytics to targeting=true impressions when you measure rule performance, and log default impressions separately as a config-coverage signal.

Context: Official docs (CloudBees, impression handler): documents the analytics gotcha: an impression with targeting=false means the flag served its default value with no active targeting rule, and treating it as a targeted exposure corrupts experiment analysis.