(Beware of) Procrustean solutions in data science

• From Wikipedia: A Procrustean solution is the undesirable practice of tailoring data to fit its container or some other preconceived structure. In a Procrustean solution in statistics, instead of finding the best fit line to a scatter plot of data, one first chooses the line one wants, then selects only the data that fits it, disregarding data that does not, so to "prove" some idea. It is a form of rhetorical deception made to forward one set of interests at the expense of others. The unique goal of the Procrustean solution is not win-win, but rather that Procrustes wins and the other loses. In this case, the defeat of the opponent justifies the deceptive means.

This happens more often than we realize in the data analytics area. Minds are already made up, investments have been made and "analytics" is used to provide final sign-off that this was their best solution. My aim in 2018 is to self-check my approach to see if I am part of the problem here.