How many users does it take to validate a product?

You can validate a design for usability with as little as 2-5 users. By 3 users you usually see the same pattern, where people go to first or things that are confusing. Validating for desirability (would people go out of their way to use this) would take a larger sample a probably a combination of experiments using A/B testing, wizard of oz and marketing research and UX research too to understand product/market fit and user expectations. I have to admit I have not done the latter for my own project which definitely makes the launch harder and a bit more of a big experiment - and I say that some what ashamedly as I am a UX designer. I have done usability tests and did some UX surveys through instagram for my app which helped me understand how I should market it and helped me curate a better brand voice that is genuine and more human.

the mix mattered less for us than what we asked people to do. we had a customer sign and pay and then never open the product once, so the written feedback looked great and the actual usage was nothing. after that i stopped counting anything people could give me for free. what worked instead was asking what they would pay before the thing existed. 987 people went through that and it came out at $18,577 in three days, and the useful part wasnt the revenue, it was that a rule picked the number instead of me picking the flattering one. your developer finding tracks with mine. developers gave the sharper critique, but neither group told me who would actually pay.

The number that matters depends on which question you are answering, and this thread has been mixing two different questions under one heading. Usability testing saturates fast, five to eight people surfaces most of the friction because interface confusion repeats across strangers quickly. Willingness to pay does not saturate the same way, because the twentieth person can still tell you something the first nineteen did not, especially if your pricing has any complexity to it.

I run a pay per case product rather than a subscription, no free trial, so validation for me was never going to be usability feedback, it had to be whether a stranger would put a card in at the point they were angriest and least patient. That number turned out to be smaller than I expected but the sample had to be real prospects mid problem, not people looking at a demo cold. A tester who is calm and being helpful gives you interface feedback. A tester who is stressed and has money on the line gives you pricing feedback. They are not interchangeable and testing the wrong group against the wrong question is the more common mistake than testing too few people.

I think first 100 users answer a lot of the question but after that focus on customers

Quality over quantity every time. Instead of chasing dozens of casual users, I'm currently validating my SaaS with a single live test-bench—a brand-new local pizzeria. Fixing real pain points in a live, high-stakes environment gives way clearer product feedback than 50 survey responses.

It depends on which of three cases you are in.

If you are building a variation on something that already exists, and you are a heavy user of that something, the number is one. You. You already know exactly where it hurts, because you hit it every day. Other people's opinions at that stage mostly add noise.

If a similar product exists but you have never really used it, you are missing that knowledge. Find the heaviest user of it among the people you know, and build until that one person is satisfied. When they are, you can ship.

The usual advice, talk to thirty or fifty people first, applies only to the third case: something with no existing equivalent, where nobody, including you, knows what the shape should be. That case is much rarer than it sounds.

I'd add one column to the feedback notes: did the tester encounter the problem naturally, or only because the test asked them to? For an extension, someone completing a guided task is useful usability feedback, but someone choosing it during their normal work answers a different question. A simple follow-up could be: 'Tell me about the last time you needed it. What did you do instead, if you didn't use it?' That also gives non-technical users something concrete to describe without asking them to critique the interface.

This is such a real founder problem. Early on, I’d rather have 5 people who match the pain than 50 random testers who say ‘looks good’. At Uxia we’ve found a useful rhythm: synthetic users for fast directional checks, then a few real users to see what actually happens in the wild. The synthetic pass helps us catch obvious UX friction before we spend scarce founder time on calls, but it never replaces the messy human feedback. Your point about watching behavior is spot on, and I still need to remind myself of it pretty often.

I love doing it on a call. I think the call makes things so much easier to get loads of good quality real feedback

Totally agree. As a founder, I’ve found a few honest sessions can teach you more than a pile of “looks great” reactions. In B2B, the right user exposing one confusing step is gold. We’re using Uxia for the real sessions and synthetic AI testing for the obvious flows, then the humans catch the weird stuff. That mix feels way more usefull than chasing a magic number.