Is hyper-parameter optimization a cherry on top or one of the core value propositions? IME hyper-parameter optimization doesn't require much in terms of implem

Is hyper-parameter optimization a cherry on top or one of the core value propositions? IME hyper-parameter optimization doesn't require much in terms of implementation effort (e.g. [1]), but requires compute. I would be surprised if a professional ML/DS user were to seriously consider paying for the implementation of the optimization. [1] https://people.eecs.berkeley.edu/~kjamieson/hyperband.html https://people.eecs.berkeley.edu/~kjamieson/hyperband.html