A data science project can offer a lot more than just a machine learning model. Understanding how the business processes are represented in the data, and what a model can learn from the data is of equal importance as the model predictions. Some of the hidden insight in your data can only be uncovered when the story behind the predictions of the model is understood. This can only be achieved with additional tools (i.e., algorithms). Many new tools for that purpose are currently being developed in the machine learning community.  
In this Blog post we introduce a relatively new algorithm for model interpretability, and for which our data science team has just released a Python implementation (PyPi and github).