The very first thing I would do is clearly define the problem or objective so I have a solid direction. Second, I would explore the data and become more familiar with it. This is extremely critical especially if I am working with a new set of data. Next, I would prepare the data for modeling. This entails data validation, detecting outliers, treating missing values, etc. With those steps completed, I would begin modeling the data until I discover the most significant or valuable results. Lastly, I would implement the model and track my results. As I'm sure you are aware, this process could vary slightly based upon the type of problem and the data and tools available.