Reading Funnels Without False Confidence

A conversion funnel that mixes anonymous and authenticated identities will invent drop-offs that never happened—and hide ones that did. Start by stating whose journey you are counting and when identity is considered continuous.

Exclude internal traffic, synthetic monitors, and known QA accounts before presenting percentages to leadership. Small apps feel these distortions most; a handful of test purchases can move a weekly conversion rate by whole points.

Compare step-to-step rates across releases, not only absolute counts. A traffic spike can improve raw completions while masking a worse experience. Pair funnel views with a simple satisfaction or task-success signal when you can.

Finally, resist the urge to optimise every step at once. Pick the largest recoverable leak that product can influence this quarter, ship one change, and re-measure with the same definitions.