When your features have different ranges, you want to bring every feature under a single hood so that one of the features doesn't get assigned more weight.
Read the latest post for more details.
How to Normalize distribution for Machine Learning