Why averages hide the app behaviour you need to see

The median journey is useful shorthand, but product decisions often live in the sequences and exceptions it smooths away.

Averages answer a different question

A completion rate tells you how often an outcome occurred across a population. It does not tell you whether a person hesitated, recovered from an error, repeated a step or changed direction. Those details belong to the instance: one bounded experience, in order, with its context intact.

Look for shapes, not anecdotes

Instance-level work is not an invitation to cherry-pick unusual sessions. Start with a defined population, establish recurring sequence shapes, then inspect representative examples and exceptions. The combination keeps the analysis grounded while preserving the texture that aggregates remove.

Use the smallest useful unit

A session is not always the right boundary. A checkout attempt, upload, search task or support interaction may cross sessions—or several tasks may happen inside one. Define the instance around the product question, document its start and end rules, and test those rules against real traces.

A useful next step: choose one active product question and write down the behavioural instance that would provide meaningful evidence. If its boundaries are unclear, resolve that before adding another chart.