Before starting to collect data, we should make sure that the goal of the project, the business problem, is completely understood. As it will guide us to what data sources to look into, and where sufficient domain knowledge and expertise is also required. For example, in the previous chapter, our goal was to predict future prices of the DJIA index, so we collected data of its past performance, instead of past performance of a European stock; in Chapter 5, Click-Through Prediction with Tree-Based Algorithms and Chapter 6, Click-Through Prediction with Logistic Regression, the business problem was to optimize advertising targeting efficiency measured in a click-though rate, so we collected click stream data of who clicked or did not click on what ad in what page, instead of merely what ads were displayed on what page.