“Five to eight users will usually surface most usability problems; after that, you keep seeing the same issues. For reliable numbers, like success rates or SUS scores, I aim for thirty to fifty in an unmoderated test, or a hundred-plus responses for surveys.”
“Five users will catch about 85% of usability issues, which is usually enough to start. I have done the worst-case version: five 45-minute interviews, recorded short clips, and used those clips to get the next round funded.”
“I used to reach for Nielsen's five-user rule until I ran a DHS-funded resilience study with 180 participants and saw how fast that falls apart when your users are not homogenous. For me the number is a function of saturation, not a rule: I stop when new sessions stop changing the design, which usually lands between five and fifteen for qualitative work, and much higher when I am measuring behavior or trust.”
“Five users is the number I’ve used for years, following Jakob Nielsen’s rule, and it’s enough to surface the obvious usability problems fast. I’ll run a cohort of five, fix what breaks, then run another five, rather than blowing the budget on one big group.”
“For a usability test, I’ve found five users will usually catch most of the big interaction problems, but the real number depends on what you’re trying to learn and how many distinct user groups you have. When I tested Momentum with nurses across three roles, I needed more than five because orientees, preceptors, and leaders were facing completely different workflows.”
“I usually stop seeing new patterns after about five users for a given task, but I rarely test with fewer than eight when the workflow is complex. At TruckX we ran a pilot with eight drivers and that was enough to cut task completion time by 61% and complaints by 73%, though I still treated it as a first pass, not a final answer.”
“I treat the classic Nielsen five as a starting point, not a contract. For VillageNCD, our quarterly contextual inquiries with ASHAs, ANMs and medical officers mattered more than hitting a number, because the real signal came from watching the same workflow break across three different cadres.”
“At the Chinese Academy of Sciences I was working with Taobao behavior data at scale, but when I later ran user research for EdTechLive’s educator tools I stopped caring about a magic number and started recruiting until the same problems kept showing up. For a usability test, five users usually surface the biggest issues; for generative research, I keep going until the patterns repeat and the team can act on them.”