Because there's no solid link between inputs and outputs. For most management techniques it's almost impossible to measure the output of a given management function. It's unfalsifiable. Like any other domain of human endeavor when operating in an unfalsifiable space you can neither confirm nor debunk a particular technique. The only thing you can do is _assert_ and flit from one fad to the next.
The paradox here is that while management seems to be required for corporate structures to function properly so much of it is fluff or outright BS. I have to wonder if a properly trained ML algorithm could replace 90% of management decisions by cutting out the cruft and ambiguity.