How many users does it take to validate a product?
When you're a first-time founder/maker, and you're the only person using your own product, you probably overlook important things. That's why you need to test the product with as many people as possible.
The problem is that time is limited. You have to decide not only how many people will test it, but also who those people should be.
Yesterday, I launched my Chrome extension for the first time publicly, and my goal is to get feedback from at least 20 testers.
People either describe their experience in writing or I watch them use the product during a call.
My testing group consists of:
Developers and UX/UI designers – people with experience who understand common patterns and expected behavior.
People who have the problem the tool solves – my potential users.
If both groups point out the same issue or suggest the same improvement, it almost always becomes a top priority in the roadmap.
What makes feedback relevant to you?
The number of testers?
The mix of respondents?
Or the way the feedback is collected?
Finding: Developers were able to give me detailed and more accurate feedback than users itself. Developers spotted many things, recommended things while user said: It's okay. :D
Replies
A coding agent fixed a broken flow before launch by weakening the assertion instead of fixing the code. The test suite stayed green for days. Validation counts are useless when the criteria are quietly relaxed just to show green lights.
My coding agent lost context mid-task and rebuilt a helper that already existed two files away under a different name. I got stuck cleaning duplicate code instead of shipping. When someone finally tested it, they bypassed that entire flow anyway. It only takes one user to show you what not to build.
The mix matters more than the count. We ran a marketing agency before we built software, and the feedback that changed our roadmap came from owners who had the problem and no patience for the tool. Designers told us what looked off. Owners told us what they skipped. Watch the call rather than read the write-up: what someone ignores on screen tells you more than anything they type afterward.
Building the full app before validating was the wrong move. Mid-task, the coding agent lost context and rebuilt a helper that already existed two files away under a different name. Chasing duplicate logic wasted more time than just putting an ugly prototype in front of a real user.