Outlier detection is important for a couple of reasons. First, it allows you to learn a lot about the extremes in your data. Typical data is usually easy to explain. If you have many values of a certain category, it is usually easy to track down an explanation. It is the extreme values that can add additional insight beyond the typical, or identify faulty processes which can be fixed.
Additionally, outliers have a profound effect upon some algorithms. In particular, regression methods can be biased by the presence of outliers and lose power because of them.