Local versus non local maximum
Imagine we are a company looking to maximize our profit with a terrible hand draw graph. Let’s say the profit is 1 million. The initial goal is to optimize the existing system to double the profit to 2 million.
See A/B testing.
But that will be a local maximum. Another much greater optimum exists, but that will require very different actions to get to it. It requires the company to think and act in a very different way.
But the risk exists in trying to find the maximum that the company instead gets into a non profitable state - I.e. bankrupt.
Â
SeeA/B testing.