Ship the ugly version

Ship the ugly version to one person who will tell you the truth.

Not to an audience, not to a launch list, not to anyone who has a reason to be kind about it. One person, chosen because they will say the thing that is actually wrong rather than the thing that is easy to say.

Every product I have built took a shape in my head over weeks that had almost nothing to do with the shape it took in someone else's hands in ten minutes. Not because I was careless. Because I knew where everything was, and they did not, and there is no amount of thinking that substitutes for watching a person fail to find the button.

The version you are embarrassed by is the one that teaches you something. The polished one only confirms what you already believed, because by the time it is polished you have committed too much to hear anything else.

There is a cost to this and it is real: showing rough work to someone whose opinion you care about is unpleasant, and the instinct to wait one more week is not laziness, it is self-protection. The week never helps. Whatever they were going to find, they were going to find, and finding it a week later means a week of work built on top of it.

Pick the person carefully, though. This only works with someone who is both willing to be blunt and close enough to the problem to be right. Blunt and wrong is worse than nothing, because you will spend a week reacting to it. The best reader I have is someone who says almost nothing and then asks one question I cannot answer.

And ask them to do a task, not to give an opinion. Opinions are cheap and generous. Watching somebody try to do the thing the product is for, in silence, is the only version of this that has ever changed what I built.

Silence is the hard part. Every instinct says to narrate, to explain that this bit is temporary and that bit is going to move. The narration is what destroys the test, because once you have explained the interface they can no longer fail to understand it, and the failure was the entire thing you came for.

Write down what they got stuck on before you write down what you plan to do about it. The two get mixed up within about a minute of the session ending, and the fix you invent on the spot is almost never the one you would choose an hour later.