All articles
8 min read

Guerrilla Usability Testing in India: Street or Recruit?

Guerrilla usability testing suits simple screens for a general audience; recruit instead when your users are a small share of passers-by, because the number of people you must approach grows fast.

By Jitendra Kumar Nirala, Founder · Updated 10 Nov 2026

A researcher watches a woman try an app on a smartphone beside a tea stall on an Indian street.

Key takeaways

  • Use guerrilla testing for simple flows aimed at almost everyone. Recruit when more than about 50 approaches would be needed to get five sessions with real users (our rule of thumb, worked out below).
  • Five testers finding roughly 85% of problems holds for comparable users. For a mixed street crowd, plan for more people or a screened sample.
  • Tell people what you record, why and who will see it before you record, and ask the venue manager before you set up.
  • Measure your own approach and yes rates in the first 20 approaches, then decide whether to carry on or recruit.
In this article
  1. 1What is guerrilla usability testing, and what can it tell you?
  2. 2How many people does guerrilla testing need, and do five really find 85% of problems?
  3. 3How many strangers will you need to approach to get five useful sessions?
  4. 4Should you test on the street or recruit participants?
  5. 5What should you say, ask and write down on the street?
  6. 6What consent do you need to record people in a guerrilla test in India?
  7. 7When should you move from guerrilla testing to recruited sessions?
  8. 8Frequently asked questions

The standard instruction is short: stand in a café, stop five people, watch where they get stuck. Most of the trouble hides in the word "people". Five strangers who happen to be in a café are not five users of your product, and that gap decides what the test can tell you.

Guerrilla testing is a good method with a narrow range. It is fast and costs almost nothing, and it finds obvious problems in a simple design before you pay for development. Crowds in India also differ by city and by language, so a café in one city can tell you little about users in another.

This page gives you the arithmetic for deciding, a script and consent line you can use tomorrow, and a plain rule for when to stop using the street and move to recruited participants. It also covers consent under India's Digital Personal Data Protection Rules, 2025.

What is guerrilla usability testing, and what can it tell you?

Guerrilla usability testing is a short, informal test where you ask people in a public place to try a prototype (a draft version people can try before it is built) or live product, and you watch where they hesitate, fail or misread. "Usability testing" means watching realistic tasks being attempted. "Guerrilla" means the participants are whoever you can stop, with no screening and no booking.

It tells you whether an ordinary person can get through a simple flow without help: finding a button, reading a label, completing a checkout. It does not tell you whether people want the product, and it cannot speak for a specific group unless that group is the crowd you are standing in.

Nielsen's 1994 case for discount usability methods (cheap, imperfect tests beat none) is the logic behind street testing. That link is our inference. Treat it as a quick check that catches the obvious, not as proof.

How many people does guerrilla testing need, and do five really find 85% of problems?

Five people find roughly 85% of the usability problems on average, but only when they are comparable users of the design. Nielsen and Landauer's average is that one tester finds 31% of the problems, which gives the formula: share found = 1 − (1 − p)^n, where p is the share one tester finds and n is the number of testers.

The average hides wide variation. Laura Faulkner drew repeated random groups from 60 test users (Behavior Research Methods, 2003). Groups of five averaged about 85%, but the range ran from about 55% to nearly 100%. Groups of ten averaged about 95%, with the weakest near 82%.

Nielsen advises about 5 users for a single user group and 3 to 4 per group when you have several groups, and at least 20 for quantitative studies. So a street crowd that only partly resembles your users should lower p. Our model, with p values we assumed for partial matches:

Chance one tester finds a given problem (p)3 testers5 testers8 testers10 testersTesters for 85%Testers for 95%
31% (Nielsen's average, comparable users)67%84%95%98%69
20% (our assumption, partial match)49%67%83%89%914
10% (our assumption, weak match)27%41%57%65%1929

Nielsen rounds the formula's 84% to 85%. The 20% and 10% rows are illustrative, not measured. The point holds either way: the weaker the match between passers-by and your users, the more people you need for the same coverage.

How many strangers will you need to approach to get five useful sessions?

Divide the sessions you want by two shares: the share of passers-by who are really your users, and the share of those who say yes. Approaches needed = 5 ÷ (fit share × yes share). This is our model, and the rates below are assumptions for illustration.

Commuters stream along a footpath while only a few stop to talk to a researcher holding a tablet.
Most people walk past, so the few who stop show why the approaches needed climb so quickly.
Share who fit your userShare who say yesApproaches for 5 sessions
50%30%33
50%20%50
20%30%83
20%20%125
10%20%250

Our rule of thumb: if your own rates mean more than about 50 approaches for five sessions, recruit. Measure it instead of guessing. Tally your first 20 approaches: how many fit (ask one qualifying question first) and how many agreed. Our guide to incidence rate shows the same screening arithmetic for larger studies.

Language narrows the crowd further. In the Internet in India 2024 report by IAMAI (Internet and Mobile Association of India) and Kantar, 57% of urban internet users prefer regional-language content. Our inference: a prototype in English may fit fewer passers-by than you expect.

Should you test on the street or recruit participants?

Test on the street when the audience is almost everyone and the flow is simple; recruit when your users are a minority, when the screens are sensitive, or when you need recordings. Use this table, which is our recommendation.

Your situationStreetRecruit
Simple flow for a general audience (an e-commerce checkout, a sign-up)Good fitNot needed
A niche user (small-business owners, frequent flyers, a bank's customers)Approaches balloon, see the table aboveBetter
Users who work in a language you cannot moderatePoorBetter
Screens with money, health or ID detailsWeak: public setting, shoulder-surfing (someone reading your screen over your shoulder)Better
Two or more user groupsNeeds about 5 for one group and 3 to 4 per group for severalBetter
You need recordings, transcripts or a report for a teamHard in publicBetter
You need numbers (completion rates)Not suitable: quantitative studies need around 20 usersBetter

Our guide to UX research methods shows how many people each method needs.

What should you say, ask and write down on the street?

Say who you are, ask for five minutes, make clear you are testing the design and not the person, and ask one qualifying question before starting. Keep to one question and three tasks. These are our recommendations.

The pitch (about 30 seconds):

"Hi, I'm [name]. I'm working on an app and checking whether it makes sense. Could you spare five minutes? You'd be testing the app, not yourself. Do you [qualifying question, e.g. order food on your phone at least once a week]?"

If the answer is no, thank them and move on. If yes, continue:

"I'll hand you the phone. Please say out loud what you're thinking as you go. I'll only note what you do. Is that fine?"

The University of Edinburgh's web team tested a sign-in service on the street in 2019, over three rounds with 30-plus participants and more than three hours of footage. They hit a venue that was too quiet, a recording app they had not mastered, and "rejection fatigue", and kept going with the reminder, "Remember it's not you we're testing, it's the design."

Notes sheet (one row per task):

A researcher takes handwritten notes while a man tries a prototype on a phone in a market lane.
A short, friendly exchange where you watch and jot down notes is all the street session needs.
TaskDone alone / with a hint / failedWhere they pausedTheir words

Our guide to the think-aloud protocol has a fuller script and timing.

Pre-flight checks (each has a pass or fail)

  • Say the pitch to a colleague with a timer. Pass: under 30 seconds. Fail: longer, so cut it.
  • Do all three tasks yourself on the device you will hand over. Pass: no dead ends or crashes. Fail: fix before you go out.
  • Time a full run of the three tasks with a colleague. Pass: under 5 minutes, matching what you ask for. Fail: over 5 minutes, so cut a task.
  • Visit the venue and ask the manager. Pass: permission in writing or a clear yes. Fail: choose another.

When should you move from guerrilla testing to recruited sessions?

Move when your own tally shows too few real users in the crowd, or when the study needs a language, a city or a recording you cannot get on the street. We can recruit and run usability sessions with screened participants when street testing is not enough.

Here is how it works. You share a brief covering your objective, method, audience, geography, sample size and timeline, and we assess feasibility and send a tailored proposal and quote. We then recruit and screen participants, with identity checks before invitation, and run remote usability testing or lab-based sessions. We can recruit by city, target by demographic and household profile, run sessions in languages other than English, and supply moderators. Deliverables depend on the scope we agree before the study begins, and can include recordings, transcripts, structured data exports and reports.

Pricing is set project by project from your requirements, so we do not have a fixed rate card. See our guide to website usability testing services. Write a short market research brief first. Send your brief to support@qualfacto.io and ask about a free pilot. We reply within one business day where we can, Monday to Friday, 10:00 to 18:00 IST.

Need screened participants for your usability sessions?

Send your brief to support@qualfacto.io and ask about a free pilot.

Ask about a pilot

Frequently asked questions

Ask for five minutes and keep the run to five, with one question and about three tasks. Longer asks raise the refusal rate. This is our recommendation; the Edinburgh team also asked for five minutes.
Yes. A café or mall controls its premises, so get the manager's clear yes, ideally in writing, before you approach customers. Public pavements usually need no permit, but check the local rules.
Yes, it is a valid early-stage method. It finds obvious usability problems, but it cannot show demand or speak for a specific user group.
A small thank-you such as a coffee is enough. Offer it after the session so it does not become the reason people agree. This is our recommendation.
Test on the street only in languages you can moderate yourself. If you need sessions in other languages, we can run recruited sessions in languages other than English. Language also lowers the share of passers-by who fit.

Sources

  1. 1.Nielsen Norman Group, "Why You Only Need to Test with 5 Users" (18 March 2000)
  2. 2.Nielsen Norman Group, "How Many Test Users in a Usability Study?" (3 June 2012)
  3. 3.Nielsen Norman Group, "Guerrilla HCI" (1994)
  4. 4.Faulkner, L. (2003), "Beyond the five-user assumption", Behavior Research Methods, Instruments & Computers 35(3), 379-383
  5. 5.University of Edinburgh, "Lessons learned from guerrilla testing" (July 2019)
  6. 6.Ministry of Electronics and Information Technology: data protection
  7. 7.IAMAI and Kantar, Internet in India 2024
  8. 8.ICC/ESOMAR International Code

Keep reading