In this step, starting from the exploratory analyses we have performed and the objectives we have declared within the business understanding phase, we start to limit the possible applicable models to a limited range, therefore defining our data modeling strategy. This is a set of possible data modeling techniques we are going to apply to our data to in order to reach the results we declared within the business understanding phase. We have to admit that, especially within industries, it is common to be guided by in the choice by the most popular alternatives. For an example, we can look at the logistic regression for the modeling of the customers default events.

This is not per say a real problem, since probably a popular technique is popular because it is showed to be effective; nevertheless we should open up our mind with more organic research, driven by technical literature and discussion with peers.

But research of what? We should focus on some kind of mix of these two components:

• How similar problems were solved in the past
• Any emerging techniques which, given the type of data we have and the question we want to answer, could fit our purposes