Sixth law: Data modeling always fails, given enough time.
This means that the model either evolves, or restricts the actions of the people using the model.
A data model has to balance flexibility and efficiency and you can’t have both extremes. I have implemented extremely flexible data models and lost the efficiency and very efficient data models that lacked the flexibility.
At work we had to do a lot of work recently because the data model we had been using was no longer sufficient for the reality it should reflect.
Fortunately we are personally and politically determined to evolve our model to reflect reality, rather than to restrict reality to fit our model. It is often one of the advantages that comes from doing the work in-house.
Henning's blog 