In Praise Of Heuristics – or Saving TOGAF® From Its Friends

By Stuart Boardman, Senior Business Consultant, Business & IT Advisory, KPN Consulting and Ed Harrington, Senior Consulting Associate, Conexiam

As the world’s best known and most used Enterprise Architecture (EA) framework, it’s quite reasonable that TOGAF®, an Open Group standard, attracts criticism from both outside and within The Open Group membership. We would like to discuss a particular class of criticism and even more about the thinking behind that.

This criticism states that TOGAF is neither rigorous nor scientific and that any good EA framework should be both of those things. Now we don’t know anyone who wouldn’t agree that TOGAF could be more rigorous about some things and that’s one of the areas highlighted for attention in the next version of TOGAF.

But, “scientific”? There’s the rub. What do we mean by scientific?

Machines, Nature and Enterprises

What these critics promote is a method, which for any given enterprise, under identical conditions will always deliver the same, “correct” result – regardless who executes the method, as long as they follow the rules. This approach depends on a very 19^th/20^th Century mechanistic view of an enterprise.

We agree that an enterprise is a system. Mechanical systems behavior is generally predictable. If you get the equation right, you can predict the behavior under any given set of conditions with an accuracy of (to all intents and purposes) 100%. So, if an enterprise were a machine, you could come up with a method that meets this requirement.

Natural and environmental systems do not, in general, behave predictably (leaving trivia like Pavlov and his dogs out of it). There is room for discussion for any one system under consideration as to why this is. It could just be because there are so many variables that we can’t capture all of them at one instant in time (i.e. they are highly complex) or because the system is chaotic (i.e. extremely sensitive to initial conditions) or even stochastic (i.e. we can only establish a probability for a particular outcome) – or possibly a mixture of those things.

A major aspect of enterprises is that, to a considerable extent, they are made up of people, individually and in groups. Each with their shifting perceptions of what “good” is. In fact even a single organization behaves more like an organism than like a machine (note: we are not claiming that organizations are organisms).

Especially important is that enterprises function within wider ecosystems in which external factors like resource availability, innovation, competition, customer loyalty, legislation and regulation (to name but a few) constantly affect the behavior of the enterprise. To reliably predict the behavior of the enterprise we would need to know each and every factor that affects that behavior. Complexity is a major factor. Do we recognize any existing enterprises that do not conform to this (complex) model?

Science and Uncertainty

Enterprises are complex and, we would argue, even chaotic systems. Change the initial conditions and the behavior may be radically different (a totally different equation). A real scientific method for EA would then necessarily reflect that. It would deliver results, which could continue to adapt along with the enterprise. That requires more than just following a set of rules. There is no “equation”. There may be a number of “equations” to choose from. Some degree of experience, domain knowledge and empathy is required to select the most adaptable of those equations. If the world of software architecture hasn’t yet determined a formula for the perfect agile system, how can we imagine the even more complex EA domain could?[1] Any such method would be a meta-method. The actual method followed would be an adaptation (concretization/instantiation) of the meta-method for the system (i.e. enterprise) under examination in its then specific context.

So even if there is an EA method that delivers identical results independent of the user, the chances they’d be correct are…well, just that – chance. (You probably have a better chance of winning the lottery!). The danger of these “scientific” approaches is that we kid ourselves that the result must be right, because the method said so. If the objective were only to produce a moment in time “as-is” view of an enterprise and if you could produce that before everything changed again, then a mechanistic approach might work. But what would be the point?

What Really Bothers Us

Now if the problem here were restricted to the proponents of this “scientific” view, it wouldn’t matter too much, as they’re not especially influential, especially on a global scale. Our concern is that it appears TOGAF is treated by a considerably larger number of people as being exactly that kind of system. Some of the things we read by TOGAF-certified folk on, for example, LinkedIn or come across in practice are deeply disturbing. It seems that people think that the ADM is a recipe for making sausages and that mechanistically stepping through the crop circles will deliver a nicely formed sausage.

Why is this? No TOGAF expert we know thinks TOGAF is a linear, deterministic process. The thousands of TOGAF certified people have a tool that, as TOGAF, itself in chapter 2.10 states: “In all cases, it is expected that the architect will adapt and build on the TOGAF framework in order to define a tailored method that is integrated into the processes and organization structures of the enterprise”.

Is it perhaps an example of the need so many people have to think the whole world is predictable and controllable – an unholy fear of uncertainty? Such people seek comfort and the illusion of certainty in a set of rules. That would certainly fit with an outdated view of science. Or perhaps the problem is located less with the architects themselves than with management by spreadsheet and with project management methodologies that are more concerned with deadlines than with quality? Less experienced architects may feel obliged to go along with this and thus draw the wrong conclusions about TOGAF.

The Task of Enterprise Architecture

Understanding, accepting and taking advantage of the presence of uncertainty is essential for any organization today. This would be true even if it were only because of the accelerating rate of change. But more than that, we need to recognize that the way we do business is changing, that agile organizations encourage emergence[2] and that success means letting go of hard and fast rules. Enterprise architects, to be useful, have to work with this new model, not to be risk averse and to learn from (shared) experience. It’s our responsibility to help our enterprises achieve their strategic goals. If we turn our backs on reality, we may be able to tick off a task on a project plan but we’re not helping anyone.

A good EA framework helps us understand what we need to do and why we are doing it. It doesn’t do the thinking for us. All good EA frameworks are essentially heuristics. They assemble good practice from the experience of real practitioners and provide guidance to assist the intelligent architect in finding the best available solution – in the knowledge that it’s not perfect, that things can and will change and that the most valuable strategy is being able to cope with that change. TOGAF helps us do this.

[1] For more on complexity and uncertainty see Tom Graves’s SCAN method.

[2] See, for example Ruth Malan and Dana Bredemeyer’s The Art of Change: Fractal and Emergent

Stuart Boardman is a Senior Business Consultant with KPN Consulting where he leads the Enterprise Architecture practice and consults to clients on Cloud Computing, Enterprise Mobility and The Internet of Everything. He is Co-Chair of The Open Group Open Platform 3.0™ Forum and was Co-Chair of the Cloud Computing Work Group’s Security for the Cloud and SOA project and a founding member of both The Open Group Cloud Computing Work Group and The Open Group SOA Work Group. Stuart is the author of publications by KPN, the Information Security Platform (PvIB) in The Netherlands and of his previous employer, CGI as well as several Open Group white papers, guides and standards. He is a frequent speaker at conferences on the topics of Open Platform 3.0 and Identity.

Ed Harrington is a Senior Consulting Associate with Conexiam, a Calgary, Canada headquartered consultancy. He also heads his own consultancy, EPH Associates. Prior positions include Principle Consultant with Architecting the Enterprise where he provided TOGAF and other Enterprise Architecture (EA) discipline training and consultancy; EVP and COO for Model Driven Solutions, an EA, SOA and Model Driven Architecture Consulting and Software Development company; various positions for two UK based companies, Nexor and ICL and 18 years at General Electric in various marketing and financial management positions. Ed has been an active member of The Open Group since 2000 when the EMA became part of The Open Group and is past chair of various Open Group Forums (including past Vice Chair of the Architecture Forum). Ed is TOGAF® 9 certified.