Good question. I think there are two parts there. How do we arrive at the place where we form the hypothesis? Usually in my case through a sketching or collaborative meeting where we look at research, context, data, insights and try to think about the signals and metrics which will help us measure the change we are hoping will happen. In terms of the cost of experimentation and potential business value from the output, that should be taken care of by prioritisation (there are various schemes but none is as good as a company specific flavour).