Sunday, 21 March 2010

Zachman: Not a Point of Departure for an Enterprise Methodology

I have spent the last few days reviewing the most contemporary Enterprise Architecture (EA) frameworks. As I am trying to establish a common description of the methodology behind EA, my goal is to sketch the intersection of methodological assumptions these framework impose on organisations. However, the analysis is still on a very high level of abstraction as my goal is a cross-framework conceptualisation. Of course, every framework has its own domain of application, knowledge, and assumptions, and the analysis is fully aware of that.

People usually refer to the Zachman Framework when addressing EA's point of departure. John Zachman published his famous first words on enterprise integration in the IBM Systems Journal publication A Framework for Information Systems Architecture in 1987, and he is still regarded by many academics and practitioners as the father of the discipline itself. In 2009 I even had the pleasure of following John Zachman presenting an updated version of his framework, and it was definitely inspiring and valuable. Besides being the first to propose an important quantum leap in the integration of business and technology, Zachman is also an excellent speaker and entertainer.

However, my analysis does not depart from Zachman's original writings, and that is with a specific reason. Zachman has continuously emphasised that his framework concerns structure and taxonomy, and in general not process or methodology. The Zachman Framework is specifically focused on representing a static state, taxonomy, or blueprint of an enterprise, but not how the enterprise has grown the architecture over time -- no development methodology is prescribed. It is merely a snapshot or slice of the enterprise at a certain point in time. As Zachman puts it: "The Zachman Framework is a schema. [...] More specifically the Zachman Framework is an ontology [...] The Zachman Framework is not a methodology." Zachman's classification structure is very useful for systematically classifying and describing the current state (or future state) or an architecture, but in my opinion he poses some fundamental erroneous assumptions about methodology and ontology -- and manages to represent it in a misleading way:

- The definition of an ontology is (according to Wikipedia): "the philosophical study of the nature of being, existence or reality in general, as well as the basic categories of being and their relations." Let us for a moment assume that it is the second sentence that Zachman focused on when using the word ontology in his writings. Ontology is in the first place quite a big word (or efficient discourse) to use when describing one out of many frameworks, and I agree with Dave Snowden when he postulates that the word taxonomy is a much better concept in the context of enterprises.

- Given that ontology concerns the "basic categories of being and their relations", Zachman must assume that a being is already in place, and that this being is involved in the creation of the relations. Being assumes a process, the creation of something -- taking part in the world -- or maybe less dramatic: the organisational field -- over time. But if Zachman's schema only concerns the organisation at a single point in time, how can it truly represent the inherent causality between the objects it claims to classify in an efficient way? A complete description of the objects and their interdependencies within an enterprise must assume a certain level of context within time and space. A snapshot for putting the enterprise in a context is simply not enough. And as ontology assumes both causality and being, Zachman's framework simply is not an ontology. Similarly, Mendeljev's periodic table (which Zachman often uses as an powerful analogy to his own concepts) not only describes the static composition of atoms (implies structure), but also how atoms evolve over time (implies process) as they wobble between a stable and unstable state. Again, Zachman's use of the term ontology is insufficient.

In short: the word ontology is simply too strong a discourse for a classification system, and a snapshot classification is again not sufficient for efficiently describing an organisation without loosing important information. How and why the organisation is in its current state are in my opinion two major prerequisites that need to be determined before we can describe an optimal future state, and the Zachman Framework is in a conscious manner sawing off the branch that its own reality check with methodology sits on.

In my search of enterprise methodologies, I have decided to take GERAM - The Generalised Enterprise Reference Architecture and Methodology -- as my point of departure. Peter Bernus, one of the authors behind the framework -- presents it as a meta methodology for describing reference architectures: “Potentially, all proposed reference architectures and methodologies could be characterized in GERAM”-- whilst emphasising the framework's ultimate purpose of methodological unification:

“GERAM is intended to facilitate the unification of methods of several disciplines used in the change process, such as methods of industrial engineering, management science, control engineering, communication and information technology, i.e. to allow their combined use, as opposed to segregated application.”

The IFIP-IFAC task force certainly created the first real stepping stones towards a true ontology and methodology for formally characterising and representing an enterprise in context. After all, that requires an inclusion of both structure and process, and these cannot be separated without leaving out important knowledge about the enterprise itself.


  1. Update:

    I received a nice reference (thanks to John Gotze) to an actual comparative analysis of core EA frameworks:

    It is written by Roger Sessions, the creator of ObjectWatch and the SIP framework.

  2. Ontology is interesting because it introduces the concept of relations into taxonomy. One of the most interesting relation is the identity relation which will help clarify category boundaries which blur and confuse many framework users.

    Your point about snap shots not being sufficient for EA is very true.

    I would suggest also considering VPEC-T to compliment your framework with process and social oriented architectural concerns.

    May I suggest reading: Women, Fire and Other Dangerous Things ( for a in depth discussion of classical and non-classical categories?