How to read drop-off between two screens without guessing

When step A shows 10,000 people and step B shows 6,400, the gap looks obvious. The harder part is deciding whether those 3,600 left the app, stalled on a slow load, or took a side path your funnel never named.

Start by fixing the clock. Compare the same cohort window for both steps, then check median time between A and B. If the median sits far above your product’s usual interaction pace, treat “delay” as a candidate before you rewrite the screen.

Next, sample a modest set of full paths — not only the exits. Look for people who skipped B entirely yet still reached a later success event. Those paths tell you whether B is required or merely common.

Only then write hypotheses. Label each as exit, delay, or bypass. That small vocabulary keeps conversion path reviews from turning into a debate about taste.

In Bangkok teams we work with, this habit shortens readout meetings: everyone argues about the same three labels instead of inventing new stories for every chart spike.