Turning numbers into pictures is usually the easiest way of finding out what they mean. We're all familiar with the display of - for example - stock prices as they change with time, and have become accustomed to quickly making decisions based on what the picture is telling us. The reason for this isn't hard to find: images are easier to understand than numbers because our eyes and brains are good at working together to discern visual relationships.
In addition, a well-chosen image can represent a vast amount of numerical data - turning it into a picture is one way of reducing the amount of space that it occupies on the page. Moreover, if it's possible to interact with the picture - by, say, changing the data, or the resolution of the display - it becomes possible to "drill-down" into the dataset by focussing on a region of interest and refining the level of detail. If the display is an interactive three-dimensional one, the viewer can fly through the data by changing their viewpoint, searching for interesting features. Finally, if the user has access to the data source, they can perform "what if" experiments, making predictions of future behaviour based on past trends and relationships between different parts of the data.