Tuesday, October 21, 2008

Can an EA be "correct"?

In a recent blog, Nick Malik asserted that EA is not about building apps right, but is about building the right apps. And that the correctness of an EA was a point of view.

This was my response:

In my view, EA is not about either building apps right OR building the right apps. It is about creating a structured framework within which meaningful and collaborative dialogues can occur between the business and IT groups. The term that I use to describe this "structured framework" is Enterprise Architecture.

In most organizations, "meaningful and collaborative dialogues" between business and IT do not occur. You must ask why this is so. Is it because the business folks are overly demanding and unreasonable? Is it because the IT folks are unable to understand what business is saying?
In my view, the problem is neither the business nor IT. The problem is that EA has either not been done at all or, if it has been done, has been done incorrectly. Which brings us back to Nick's original point. Is there a "correct" way to do EA?

The answer is absolutely, yes. But the starting point needs to be an understanding of what we mean by "correct". I define "correct" as "as simple as possible". So of two EAs that both accurately define the enterprise, the more correct one is the simpler of the two. And the EA that is "absolutely" correct is the EA that is the simplest possible EA.

Why do I believe that simplicity is the raison d'etre for EA? The answer goes back to my original question. What is it that gets in the way of meaningful dialogues between business and IT? In my opinion, what gets in the way is complexity.

When IT blames the business for fuzzy requirements, IT is wrong. When the business blames IT for inability to deliver, the business is wrong. The problem is neither IT nor the business. The problem is complexity. Until complexity is solved, dialogue is impossible. Complexity is the enemy, and it is the common enemy of both IT and the business.

So can you determine if an EA is as simple as possible? Again, the answer is yes. But to do so requires both an understanding of the mathematics of complexity and a rational process for testing an EA for being "as simple as possible".

These are both topics I have covered extensively in my book, "Simple Architectures for Complex Enterprises", so if you are interested in these ideas, that is where to go. I also have a number of papers on this topic at my website, http://www.objectwatch.com/.

Best Wishes,
Roger Sessions

No comments: