Five Is the Magic Number

Definition

Testing a prototype with five carefully recruited target customers, one at a time, reveals nearly all the important behavioral patterns a team needs — watching a sixth, seventh, or eighth mostly confirms what the first five already showed. The book credits usability researcher Jakob Nielsen, who analyzed 83 of his own studies and plotted problems found against number of interviews conducted: roughly 85% of usability problems turned up after just five interviews, after which the discovery rate "despenca como uma pedra" (drops like a stone) — so the efficient move is to fix the 85% you've found and test again, rather than pay for diminishing marginal insight from a bigger sample.

In the Book

Chapter 15 sets this up with the story of publisher Nigel Newton, who in 1996 handed a manuscript rejected by eight publishers to his eight-year-old daughter Alice rather than reading it himself; her unprompted enthusiasm — "this is so much better than anything" — led him to publish it: Harry Potter and the Philosopher's Stone. The book uses this to frame Friday's five-customer test as the same kind of foresight: watching a real, uncoached reaction before committing further resources, rather than guessing what customers will think. It then walks through One Medical's clinic-lobby prototype, where seeing just two children struggle to wheel strollers past a cramped entrance was enough to identify the problem — no large dataset was needed, and when two or three of five interviewees show the same reaction, positive or negative, the book says to treat it as a real pattern. Five interviews also fit conveniently into one day with breaks between them (the book gives a literal 9:00-to-16:30 schedule), letting the whole team watch live and skip waiting for someone to compile results afterward.

Why It Matters

Nielsen's curve reframes "how big a sample do I need" from an assumption (bigger is always better) into a shape (steep early returns, then a sharp elbow), and names the point past which more data mostly buys confidence rather than new information. It also supplies a reason qualitative, small-N interviews beat larger quantitative surveys for a specific job: interviews can surface the why behind a reaction, which statistics alone cannot, and Friday's design deliberately optimizes for that "why" over sample size. This generalizes to any research or feedback process where the cost of adding another data point is real and the question is when to stop collecting and start fixing.