Reading funnels when samples are small
Early-stage products rarely have the volume for tidy statistical tests. Here is how to still learn from drop-off without inventing false certainty.
Small samples do not make funnel analysis useless—they change how confidently you should speak. A 40% drop between step two and three on 200 users is a signal worth investigating, not a headline for the board deck.
Pair quantitative drop-off with five to eight session recordings or moderated tasks on the same path. The combination usually reveals whether users misunderstand copy, wait on a slow screen, or simply do not want the next step.
Avoid slicing every funnel by every segment at once. Choose one cut that matches a real decision—new versus returning, or iOS versus Android—and leave the rest for later when volume grows.
Document what you will revisit after the next thousand users. That habit keeps early analysis honest without freezing product work until perfect data arrives.