The short version
I would rather ship something small than plan something large.
Most of what I have built started as a weekend annoyance. That is not a coincidence — it is the only reliable way I have found to tell whether a problem is real before spending three months on it.
The rest is unglamorous: talk to the eight people who have the problem, build the smallest thing that could fail, and watch what they do instead of what they say.
01
Find the annoyance
Something I keep working around by hand. If nobody else works around it, it was a preference, not a problem.
02
Build the smallest version
One screen, real data, no account system. Fast enough that I am testing the idea and not the code.
03
Watch, then decide
Two weeks of use answers the question that two months of discussion does not.