I answered a distribution problem with product work four times. Each time the reasoning was plausible and the result was a better product and identical silence.
What broke it: no product work in a week where I had not spoken to somebody who was not already a user. Most of what I built in those two years turned out to be unnecessary.