Novelty vs Utility
let's say in order to predict the churn rate of the customers you came up with 2 machine learning approaches - logistic regression and neural networks.
you know that logistic regression models will be highly interpretable and you will be able to identify the important features whereas the neural networks model, even though will give a better performance will be less interpretable since it's a black box model( it won't explain clearly why it made a certain prediction)
Which modelling technique should you ideally prefer?
Answer: You need to convey to the client which customers are leaving as well as the features that are more important for their departure. Therefore interoperability matters to them. Hence you should preferably go with the Logistic Regression Model.