Production value used to be expensive. That was the only reason it counted as evidence.
For about fifteen years, a well-built prototype was a legitimate signal. Not because users cared how it looked, but because building something that looked finished took a designer, an engineer, and several weeks of somebody's salary. If you had one, you had convinced people to spend real resources on your idea, and you had made a hundred small judgement calls in public. The artefact was proof of the effort behind it.
That inference is now broken, and it broke quickly.
Why did polish stop being a signal?
Because the cost went to zero and the meaning was never in the polish itself. It was in the cost.
Anything that works as evidence works because it is expensive to fake. A degree, a reference, a working demo, a queue outside a shop. The moment the price of producing the appearance falls below the price of producing the underlying thing, the appearance stops carrying information. That happened to interfaces some time in the last two years.
The trap is that the appearance did not get worse. It got better, and more uniform. Clean landing page, dark mode, tidy cards, a plausible AI feature, a deployed demo. Everything looks like it works. Nothing tells you whether anyone wanted it.
What is still evidence after polish collapsed?
Behaviour that costs the person something.
- Returning without being asked. Nobody opens a tool twice out of politeness.
- Paying. Not agreeing that the price seems fair. Actually paying it.
- Complaining about a specific thing. People do not file detailed complaints about products they do not use.
- Adapting around a gap. A user who builds a spreadsheet to patch something your product does not do has told you more than any interview will.
- Asking when it will be ready. Unprompted, twice, from the same person.
Each of these is expensive to fake, which is exactly why each survived.
How can a founder tell the difference in practice?
Ask one question about any piece of evidence you are holding: what did this cost the person who produced it?
A waitlist signup costs an email address, which is nothing, so fifty of them mean roughly nothing. A friend telling you the idea is great costs them a moment of social effort in your favour, so it is worse than nothing, because it is actively biased. A stranger who came back on Tuesday without a reminder spent attention they could have spent elsewhere. That one counts.
This is also the honest reason a beautiful demo and a bad business coexist so comfortably. The demo is a statement about your team. Demand is a statement about other people. They are different claims and only one of them is about the market.
Does this mean you should ship something ugly?
No, and this is where the argument usually gets misused.
Polish is not harmful. It is just not informative. Build something that works well, because the people you put it in front of deserve that, and because a broken thing generates noise rather than signal. Simply stop counting the quality of the artefact as part of your case that the idea is wanted. It belongs in a different column.
The version of this that hurts founders is subtler. It is not that they think polish equals demand. It is that months of visible progress feel like months of progress, and there is no moment where anyone stands up and says the evidence column is still empty. The build is legible, daily, and satisfying. The evidence is not, so it gets postponed.
Key takeaways
- Polish worked as evidence only because it used to be expensive, and it no longer is.
- Evidence is anything costly to fake, which is why behaviour outperforms opinion.
- A waitlist signup, a compliment and a nod are all cheap, and their volume does not fix that.
- Returning unprompted, paying, and working around a gap are the signals that survived.
- Ask of every data point you hold: what did this cost the person who produced it?
FAQ
Does a good looking demo help with fundraising even if it proves nothing about demand?
Yes, and that is worth saying plainly. It shortens the conversation and shows competence. It just is not an answer to the question of whether anyone wants the product, and problems start when the two get quietly merged in the same slide.
How many real users do I need before it counts as evidence?
Fewer than most people assume, if the behaviour is strong. Five people using something weekly, unprompted, for a month is a far stronger position than a thousand signups, because the five are demonstrating cost and the thousand are demonstrating curiosity.
Is user feedback on a prototype useless then?
Not useless, but it answers a different question. Feedback on a prototype tells you whether the thing is usable and understandable. It does not tell you whether it is wanted. Confusing the two is how teams end up with a well-tested product nobody buys.
If polish is free now, what is the new expensive thing?
Attention and money from people with no relationship to you. Those are the two resources that did not get cheaper, which is precisely why they still carry information.
What is the fastest way to tell if I am fooling myself?
Write down, before you look at anything, what result would make you stop. If you cannot name one, you are not gathering evidence, you are gathering reassurance.