Data Modeling: Art or Science?A practitioner recently commented: “Everyone has their biased view of what a data model is. Data modeling is art – not science. Give 6 data modelers one set of requirements and you’ll get 7 solutions all distinctively different.” My response: To me that’s a huge problem. No, ‘data’ modeling is not a science, but nor should it be an art. Actually, it should be engineering. Engineered solutions have to stand up to rigorous tests. But we lack that in ‘data’ modeling. Why? Because ‘data’ modeling is divorced from its initial business context, which is operational business communication, including business rules. You need nouns and verbs for that, and those nouns and verbs should stand for well-structured concepts. Give me a model of well-structured concepts that has been ‘proven’ by verbalizing business rules and other formal business communications and I guarantee I can come up with the best data model. I’m talking of course about concept models (sometimes called fact models).
Tags: business communication, concept model, data model, fact model, verbalization
In my experience the information found in concept models does not always have a 1:1 relationship with the data held in associated data bases. For example the method of collecting the Full Names of entities in a form by providing prescribed combinations of fields for Given Name(s), Family Names, Organisation Name, etc may not directly correspond to the way or ways the Full Name is held in a data base. There are all kinds of reasons for variations of holding the Full Name data in a data base; all, you would hope, related to business requirements, but probably unfortunately related more to the “artistic taste” and/or learned guidelines of the data modellers.
I agree fully with your statement “Give me a model of well-structured concepts that has been ‘proven’ by verbalizing business rules and other formal business communications and I guarantee I can come up with the best data model”. But I can see the look on the faces of the data base keepers when that “best data model” exposes flaws in the existing data base design. In fact I do see the look on the faces …!