Hey everyone 👋 Running into an issue and wondering if anyone's seen this before. We're running an experiment with

fpid

(a first-party anonymous ID) as the assignment attribute. Traffic is set to 100%, 50/50 split, no namespace. We're seeing that ~30% of users never get bucketed into any variation, and it's disproportionately concentrated on iOS — both mobile Safari and in-app browsers (Meta webviews) coming from our ad funnels. Has anyone else dealt with a chunk of iOS traffic — particularly from ad-platform in-app browsers — falling out of experiments entirely despite 100% traffic inclusion? Would love to know if there's a known root cause or a recommended pattern for making the assignment attribute more resilient on iOS/webview traffic. Thanks!

Hi @adamant-guitar-27978 good morning 👋 Thanks for the assisted information. At 100% traffic, GrowthBook includes every user who passes the

onboarding_name

targeting condition and has a non-empty `fpid`; if

fpid

is missing when the flag is evaluated, the user will not be bucketed. Please can you confirm whether the affected users receive the flag’s default value or are assigned correctly but missing the exposure event due to consent or analytics blocking? It would be helpful to debug this by logging

fpid

and

onboarding_name

immediately before evaluation, the returned flag source/value, and whether the tracking callback successfully sends.

@flaky-noon-11399 thx for your response, continue investigating this with my team will back with updates