Here’s another unpopular (but secretly popular) opinion: it can be perfectly reasonable to make a small product change on the strength of a little feedback. Sometimes even on expert judgment. Just keep measuring.
The stakes matter
A small, reversible adjustment with limited reach is not the same as a change to a critical workflow used by millions. Treating both as if they need identical research is not rigor. It is a failure to account for risk.
Scale the evidence
As impact, uncertainty, or irreversibility increase, so should the evidence. Look for independent signals, ideally collected in different ways. Prototype before you commit when you can.
And keep measuring after release. Validation is not a ceremonial gate you pass through once. It is an ongoing conversation with reality.
A useful question
Before choosing a research method, I like to ask: what could go wrong, how would we know, and how quickly could we recover? That tends to produce a much more sensible plan than arguing about whether the process is “proper UX.”