School taught me hypothesis, experiment, conclusion — does actual research work like that?

I want to defend the school version slightly. It is a bad description of a week in a lab and a decent description of what makes a claim count as supported: you said in advance what you expected, you did something that could have shown you were wrong, and it did not. Everything in the messy real version is in service of eventually being able to do that one clean thing. Teaching fourteen year olds the messy version first would leave them with no criterion at all.

For my first hobby project I wrote the hypothesis down, did the experiment, got a mess, and concluded I had done science badly. What I had actually done was a pilot, which is what everyone does first, and the correct response was to fix the measurement and run it again rather than to write it up as a failure. Nobody had told me that the first version of an experiment is for finding out what you are actually measuring.

It is a description of the logic of justification, presented as if it were the workflow, and those are different things. Real projects usually start with a vague dissatisfaction, wander through a lot of exploratory work that nobody will ever see, and only crystallise into a testable claim quite late, at which point you do the clean confirmatory experiment that ends up as figure one. The paper is written backwards from the result because that is the order in which the argument makes sense, not the order it happened. That is legitimate as long as the exploratory and confirmatory parts are honestly labelled, which is exactly why preregistration exists, to make the distinction visible rather than to change how anyone works.

Some rough numbers from my own corner, which is small scale lab work. A project that produces one paper typically runs eighteen months to three years, and of the ideas we take seriously enough to spend a week on, maybe one in five survives contact with a pilot experiment. The single figure in the paper that looks like a clean result is usually the fourth version of that experiment, with the first three having taught us that our measurement was picking up something else. None of that is hidden, it is just not interesting enough to write down.

Notice something, poke at it, work out what would have to be true, then do the clean version. The diagram is the last step drawn as if it were all of them.