What people actually do.

The user in the room during planning is a strange creature. They have read the page. They remember what they did last week. They understand the difference between an account and a profile. They are not doing anything else. They want to explore the product.

The user outside the room is doing this on a phone, at 9pm, with one thumb, half-watching something else, having arrived from a link they do not remember clicking, in a language that may be their second. They are not exploring. They came to do one thing and they are estimating, constantly and unconsciously, whether this is going to be worth it.

Almost every avoidable product failure I have seen comes from designing for the first person and shipping to the second.

People are not lazy. They are triaging.

“Users won't read” gets said with a sigh, as though it were a character flaw. It is not. It is a rational response to a lifetime of interfaces that wasted their attention. Skimming is a skill people developed because most text on a screen does not repay careful reading, and they are right about that more often than we are comfortable admitting.

So the instruction is not “write less.” It is: earn the sentence. A person will read carefully at exactly the moment they suspect the next click costs them something, money, time, their phone number, a commitment they cannot undo. That is when your explanation is wanted, and that is where it should be. An explanation offered before the person feels the question is noise, and being right does not make it less noisy.

Confusion does not look like confusion.

People almost never tell you they were confused. Confusion is embarrassing, and the socially available explanation is always available: I got busy, I'll do it later, I changed my mind. In analytics it looks like a clean drop-off. In a support ticket it looks like a different question. In a user interview it looks like a person being generous about a screen that quietly humiliated them.

This is why I distrust both extremes of the research religion. The numbers tell you where people stopped and lie about why. The interviews tell you a story people have already tidied. You need both, and you need to treat the tidy story as a hypothesis rather than a finding.

Hesitation is the real unit of work.

If I could instrument one thing it would not be clicks or time-on-page. It would be hesitation: the pauses, the re-reads, the field someone fills, clears, and fills again. That is where the product is actually being decided, and it is almost entirely invisible in a standard dashboard.

You can find it without instrumentation if you look in the right places. The question support answers most often. The field with the highest correction rate. The step where the same person comes back three times across two days. Each of those is a person telling you, without words, that something in the sequence asked more of them than it explained.

The hardest thing to accept

People do not want your product. They want the thing on the other side of it, and your product is the toll. This is not cynicism, it is the most useful piece of respect available to a product person, because it kills the entire category of work that assumes engagement is a goal rather than a cost.

The moment you accept it, a lot of decisions get easier. You stop trying to make the flow delightful and start trying to make it short. You stop celebrating time-in-app. You start noticing that the best version of many features is the one the person never has to see.