You do not need an audience. You need eight people who will tell you the truth.

Around company number two, I showed something I had spent five months building to roughly thirty people. Every one of them said it looked great.

It did not look great. The pricing made no sense, the first screen never said what the product actually was, and you had to click twice before anything happened. I found all of that out about a year later, from a stranger, in four minutes.

So here is the short version, in case you bounce: you do not need an audience to find out whether your product works. You need about eight people who have used it and are willing to say something unflattering to your face. Eight is roughly where new problems stop appearing. Thirty polite people will teach you less than three honest ones, and building the audience first is the most expensive way to postpone finding that out.

Why an audience is the wrong instrument

An audience is on your side. That is the entire point of an audience, and it is exactly what makes it useless for this particular job.

People who followed you because they liked your build-in-public thread have already decided they want you to win. When they look at your product, they are not evaluating it. They are being supportive, which is lovely, and which produces a stream of "this is sick, congrats!" replies that feel like validation and contain no information whatsoever.

Worse, the feedback you do get is filtered through the thing they know about you. They know what you were trying to build, because they watched you build it. A stranger does not. A stranger arrives at your landing page with no context, gets confused, and leaves. That confusion is the data. Your audience is structurally incapable of producing it, because they cannot un-know what you told them for six months.

Jakob Nielsen ran the numbers on this back in 2000 and the finding has held up better than most things from 2000: five users will surface something like 85% of the usability problems in an interface. The curve is brutal. User one finds a third of everything. By user five you are mostly watching people trip over the same three things. People have argued with that number for twenty years, mostly on the grounds that it depends heavily on task complexity, and they are right. Which is why I say eight rather than five. Eight buys you margin, and it covers the case where two of them turn out to be too polite to be useful.

Eight. Not eight hundred. The gap between those two numbers is where a lot of founders lose a year.

What eight honest opinions actually cost

You can buy this. It is a real market with real prices, and I think more founders should know what the going rate is before they decide it is not worth the trouble.

A roast from fffuel is $89. Roastd.io is $179. MakerBox is $298. Olly Meakings charges $350. These are real people writing real critiques, and having read a few, they are worth the money.

Eight of them, at the cheapest credible end, is $712. At the mid-market end it is $2,384.

That is the number I want you to sit with. Not because you should go and spend it, but because it puts a price on something most founders treat as free and then never actually acquire. Honest, structured feedback from someone with no stake in your feelings costs between seven hundred and two and a half thousand dollars on the open market. Meanwhile the standard advice is to spend six months building a Twitter following, which costs considerably more than $712 in time and delivers worse data.

I am obviously biased here, because I run a thing that arranges these trades between founders instead of selling them (so, yeah, take the framing with salt). But the arithmetic holds whether or not you use it. Two founders reviewing each other's products properly have each produced something worth a few hundred dollars, and it cost them an evening.

The part nobody warns you about

Getting the feedback is the easy half. The hard half is that most of it will be about your positioning, and you will not want to hear it.

I have watched this pattern enough times to expect it now. A founder asks for feedback on their product, braced for someone to tell them the onboarding is clunky or the mobile layout breaks. What they get instead is eight people who could not work out what the product was for, or who thought it was for someone else entirely, or who understood it perfectly and did not believe the claim on the homepage.

That is a much worse thing to hear, because it does not have a Jira ticket attached to it. "Your empty state is confusing" is a nice problem. You can fix it on Tuesday. "Nobody understands who this is for" means the thing you have been telling yourself for five months is wrong, and there is no ticket for that.

Which is why the polite-friends version of feedback is so seductive. It only ever generates the Tuesday problems.

Who this is wrong for

I would be doing the thing I just complained about if I did not say who should ignore all of this.

If you already have product-market fit and a funnel that converts, eight opinions will not help you. Your problem is volume and your instrument is analytics, not conversation. Go read your funnel.

If you are building something consumer with a network effect, where the product genuinely does not work until thousands of people are on it, then yes, you need an audience, and you need it early. Eight people cannot evaluate a social network.

And if what you actually want is encouragement, which is a completely legitimate thing to want when you are eleven months into building alone, then please go and get it. Just do not confuse it with research. Get your encouragement from people who love you and your feedback from people who do not care about your feelings. Asking one group for both is how you end up with thirty compliments and a product nobody buys.

What I would do with a week

Find eight founders building something adjacent to yours. Not competitors, adjacent. Offer to properly review their product, in writing, with specifics, in exchange for the same. Actually use the thing for an hour before you write a word, because a review written from someone's marketing site is worth close to nothing and everybody can tell.

Then ask them the only question that matters, which is not "what do you think": ask them where they first got stuck, and what they thought the product was for before they used it.

Eight answers to those two questions will be more useful than everything you learn on Twitter this year. It costs you eight evenings, which at the mid-market rate above is about $300 an evening of value produced, and unlike the roast, you get to keep the relationships.

I am still not very good at asking for this. There is a reflex, when someone offers to look at something I made, to explain it first, to get ahead of the confusion, to say "oh the onboarding is a bit rough right now, ignore that". Every time I do that I destroy the data I asked for. I am working on it. Guess I'm still human.