Most of what gets applauded in software is theater. A shiny demo. A framework migration retold as a hero story. A launch post that thanks fifty people for a feature that nobody asked for. I have sat in those rooms. I have written those posts. They feel like work, which is the dangerous part, because they are not.
The only honest measure of progress is whether the user's pain moved. Not their delight score. Not their NPS. Their pain. Something they could not do before that they can now — or something they had to do before that they no longer have to. Everything else is decoration we put on top of the outcome.
The demo is not the product
I have built features that demoed beautifully and failed to move a single metric. I have also shipped quiet, unglamorous refactors that halved the time a support agent spent resolving a ticket. Only one of those was work. The other was performance.
"The test of the engine is not that it starts; it is that it still runs after an hour."
— FIELD NOTE, WENG SENG
The hardest discipline I try to hold is letting a beautiful feature die when the evidence says it is not needed. That is not a product failure. It is the product function working. The failure is the opposite — keeping a feature alive because someone gave a good talk about it.
Rhetoric is cheap; evidence is expensive
Arguments are free. Anyone can produce one in an afternoon. Evidence costs instrumentation, a real user base, patience, and the humility to be surprised. When an argument and the numbers disagree, the argument is the thing that needs to change. It does not matter whose argument it is. It does not matter whose numbers they are.
I would rather ship something ugly and true than something beautiful and performative. The first one compounds. The second one eventually asks to be replaced by the first.