1. Steps depend on inconsistent platforms
If web and mobile emit different event names or trigger at different moments, the combined funnel may compare unlike behavior.
2. Conversion changes exactly at release time
A sharp movement aligned with an instrumentation release deserves validation before interpretation. Compare raw event volume and property completeness.
3. Users appear to skip required steps
This can reveal alternate paths, but it often points to missing events, stale sessions or identity stitching problems.
4. The window has no product rationale
A one-day conversion window can undercount a considered workflow; a thirty-day window can connect unrelated sessions. Choose it from expected behavior.
5. The denominator changes between teams
“Started onboarding” might mean account creation, first screen view or email verification. Write definitions beside the chart and resolve differences before debating performance.
Validate critical funnels with test accounts and sampled event sequences. A small quality check is cheaper than a roadmap built on broken evidence.