Getting to the root of the business problem

Some problems are easy to solve, while others prove to be much harder. One of the reasons this happens is that when a problem's symptoms appear somewhere else and after some delay, then it is very difficult to know where the problem really is. By definition, the symptoms are clearly visible—they are explicit and you can easily collect data about them. The underlying problem, on the other hand, is happening in some other department or building and is not visible because it is not causing immediate pain. Most likely, no data is being collected about the root problem, or it might be too hard to collect that data. Given the nature of ML, it is almost a given that all the data you are getting is about symptoms. If you are lucky, you might get some data about the root problem as well (although you will not know it).

One of the ways to get started is by using an old method called five whys, which basically involves asking the question...