The need for causality is best illustrated by a real life example.In the long gone, go-go days of the stock market, the Dow Jones average was highly correlated to the height of the miniskirt above the knees. This correlation was extremely good. It worked for a few years.The trouble with only a correlation is that it doesn't consider cause and effect. A model for the response has to be built up including all factors one can think of. Then data is used to try and sort out which factors are important. Leave out a factor that is really important (e.g. a war, in the case of the stock market) and the conclusions are invalid for many cases. Trying to explain something as complicated as the stock market by a single factor was naive and of course destined to fail in the long run. However, some people bet their money on it! The miniskirt is back and the stock market is not doing very well today.