Is “Data Model” an Abused Term?

One of the things I learned in writing my new book Definitions in Information Management  is that what seem to be very precise terms can be used to mean quite different things in different contexts.  In fact, terms that have a good deal of precision - and respect - can become targets of "semantic hijacking" so they end up being used for poorly defined concepts, or for  empty concepts.  It seems that this is now happening to the term "data model" .

When I first began using ERwin 20 years ago, "data model" was both a new term and a new concept.  Up to that point there had been little in terms of support by tools or methodologies for creating a precise and compete design of a database prior to its implementation.  Nor has there been much of an idea about representing the data the business deals with in the way in which the business understands it.  With IDEF1X and ERwin we got support for both physical and logical models.  And they were visual - you could print out data models and hang them on the wall.

But over the years a funny thing has happened.  The term "data model" has started to be applied directly to physical database instantiations.  I now hear people saying that "Product X has a great data model".  To which I reply "Sounds good - can you show it to me?".  I then get puzzled looks, implying that I have no idea what a "data model" is.

What is actually meant by "data model" in this usage is, of course,  not a data model we would produce using ERwin or any other data modeling tool.  It seems that what is meant is the overall features of a database, usually as understood by non-data analysts in the context of some product or application.  In these circumstances there is no model, and there is usually no intention of ever having one.  And even if one were produced, the individuals using the term "data model" would never look at it.  The problem is that the term "data model" is being used in a way that implies the benefits of a true data model are present whereas in fact they are not.  It is being used to imply methodological and engineering precision, and an investment to produce quality analysis and design artifacts.  But in reality there is no precision, no investment, and no quality artifacts. 

If we allow "data model" to become little more than a vacuous marketing term then not only will the term be devalued, but so will the standing of data modelers.  We all need to be wary of this and should push back when we encounter the term "data model" being applied to a situation where no data model exists.