The Rule of Five: The Mathematical Case for Small Prototype Tests
Type: media · article
Stage: Stage 4: Prototype Proof
Difficulty: intermediate
Five users will surface 85% of your usability problems. The sixth surfaces roughly 3% of new problems. After five, you're paying for diminishing returns. Run three rounds of five rather than one round of fifteen — each round starts fresh, after the previous round's problems have been fixed. The Rule of Five only holds within a single user segment.
Overview
Founders consistently over-invest in user testing. They recruit twelve people, run week-long sessions, and produce forty-page research reports before making a single design decision. This is backwards. The research that actually moves prototypes forward is faster, smaller, and far cheaper than most founders realize. Jakob Nielsen and Thomas Landauer proved this mathematically in 1993. Their finding has not changed. Five users is almost always enough.
The math
In qualitative usability testing, any given user has approximately a 31% chance of encountering any given usability problem. This is the average discovery rate, derived from Nielsen and Landauer's analysis of dozens of studies.
Plug that number into their formula and you find that five users will surface approximately 85% of the usability problems in your prototype. The sixth user surfaces roughly 3% of new problems. The seventh, less than that.
After five users, you are paying for diminishing returns. You are not learning proportionally more — you are confirming what you already know.
What this means for Stage 4
Run your first prototype test with five users from your target segment. Not twelve. Not twenty. Five.
Then take what you learned, fix the most significant problems, and run another five. Three rounds of five-user tests will surface more diverse and actionable insight than one round of fifteen — because each round starts fresh, after the previous round's problems have been addressed.
The caveats that matter
The Rule of Five only holds when your user group is relatively homogeneous. If your product serves two fundamentally different types of users — say, an administrator and an end user, or a B2B buyer and a daily operator — you need five users from each segment. Test them separately. Mixing segments produces contradictory data that sends you in the wrong direction.
The Rule of Five is also specifically for qualitative testing: observing what users do and why. If you need statistical reliability — measuring task completion rates, comparing two versions — you need a larger sample. Twenty to forty users minimum. But at Stage 4, you are not running statistical studies. You are watching people interact with a prototype and looking for patterns.
The practical rule
Five users. One segment. One session per round. Fix the problems you see. Run another five.
Stop when you complete a round and nobody gets lost in the same place twice. That is your signal that the core experience is working well enough to build on.