What Is Guerrilla Usability Testing?
Steve Krug's rule in Rocket Surgery Made Easy is "a morning a month, that's all we ask" - one round of testing every month, three participants a round, done in a single morning. Guerrilla usability testing is that same low-budget, low-ceremony approach: approach potential users in public spaces - coffee shops, libraries, co-working spaces - or run quick remote sessions, and ask them to complete a task with your prototype. No recruiting agency, no lab, no week of scheduling emails.
Why Three, Not Five
Krug deliberately breaks from Jakob Nielsen's better-known "five users" figure. His reasoning: three users you actually test beat five you plan to get around to. Guerrilla testing trades some statistical rigor for frequency - a small round every month, with fixes made between rounds, surfaces more problems over a quarter than one large study you keep postponing.
Where to Find Participants
Krug borrows bank robber Willie Sutton's line about why he robbed banks - "because that's where the money is" - as the guerrilla recruiting principle: go where your users already are. For a consumer product, that's a coffee shop or a co-working space. For a more specific audience, it's wherever that audience already congregates - a library for older adults, a professional meetup for domain experts.
How to Run a Session
- Introduce: "Excuse me, I'm testing a design and could use 10 minutes of your time - it's not a test of you, it's a test of the design."
- Observe: give a specific task, not a tour - "Imagine you want to book a same-day appointment. Please show me how you'd do that" - then stay quiet and let them narrate their thinking as they go.
- Debrief: ask what they expected to happen versus what actually happened. Note what they did, not why you think they did it.
Krug's fix philosophy is "the least you can do": don't try to solve every problem a session surfaces. Pick the smallest change that addresses the biggest problem, ship it, and test again next month.
When Not to Use It
Guerrilla testing is a poor fit for B2B products that need specialised domain expertise, products for vulnerable populations, or any study where statistical validity is the actual deliverable. When representativeness matters, a recruited sample beats whoever happened to be in the coffee shop that morning.
Key Takeaways
Three participants, once a month, is Krug's whole system - frequency beats scale for catching problems while they're still cheap to fix. Recruit where your users already spend time, keep tasks specific, and fix only what matters most before the next round.
