For small familiar data models common sense and experience probably suffice but for anything larger or more complex they are invaluable and should be followed wisely.
For the three Rapid Development modules I'm running, I've covered one or more of the following at both Undergraduate and Postgraduate level:
Entity relationship modelling, Normalisation, Class Diagrams and Data Dictionary.
Out of all of these Normalisation seems to be the topic which is being described in several different ways, each more unnecessarily complicated than the rest. The columnar-tabular method I was taught in industry seems to be the most effective way of teaching normalisation and I'm using my old training course material to deliver it in the class room. I will keep trying to track down an author who completes normalisation in the same way but in the meantime if anyone reading this finds any suitable papers, websites or books then please let me know...