The route between the steps
People revisit screens, change intent, pause and return. A single sequence can report completion while hiding repeated loops or costly detours. Start by listing the routes accepted by the product, not only the steps preferred by the report.
Read absence carefully
A missing event might mean abandonment, a tracking fault, an offline transition or a legitimate alternative route. Validate the instrumentation and compare neighbouring signals before assigning a behavioural explanation.
Use the smallest honest model
A useful funnel has a named audience, entry condition, completion definition and time window. Document those choices beside the result. This makes comparison possible and keeps a clean percentage from implying more certainty than the data holds.