I’ve literally never seen that happen. It’s always problem, initial hypothesis, request for data (then data is either unavailable or typically supports the hypothesis, occasionally the data doesn't and you go back to the drawing board.
Have you worked on a data team? I've seen that bs a number of times, it's how I mentally grade different managers and PM/POs.
Re the unavailable data: Smart people ask before a big change, get told what devs are missing and need to instrument/record and then leverage those new metrics for the before/after comparison. Not-smart people yolo the changes, ask for the metrics after and go whoops it's too hard or impossible to check.
