In my first exposure to business process re-engineering, our chosen consultant explained how Information Systems is “just like a factory.” We have a process for understanding the products we’re supposed to create (systems requirements analysis), another process for designing them (systems design), one for building them (systems development) and one for distributing them to our “customers” (operations). Since factory and systems processes line up one for one, IS is a factory, right?

And it is, except for everything we do. I could probably figure out a one-to-one correspondence between a factory and human courtship rituals, but that doesn’t mean human courtship is a factory either. (Let’s agree to not explore the parallels.)

In our ongoing quest to improve ourselves, many of us have tried to apply process redesign to IS, and for the most part the results haven’t been pretty. The biggest problem has been a misunderstanding about what constitutes a process. It’s a process if you do it over and over again. It’s an ad hoc task if you do it once or infrequently. Carefully designing and documenting processes for handling ad hoc tasks is pointless.

This week we continue to create our integrated IS plan, zooming in on the core processes of IS. Your first step is to identify the core processes that most need fixing and include in the plan initiatives to improve them. Improving them is important because employees only have five choices in how they get their work done. They can:

1. Improvise constantly because there is no well-defined process.

2. Slavishly follow a poorly designed process that leads to mediocre results.

3. Fight with a poorly designed process that interferes with achieving excellent results.

4. Improvise constantly, ignoring the well-designed process that would facilitate the creation of high-quality results.

5. Follow a well-designed process to create high-quality results.

It isn’t a core process unless you do it over and over again, so the process you instinctively thought of first — systems development — is rarely worth your time and attention. (You buy whenever you can, building new applications only when you’re desperate.) If you’re a typical IS organization, your core processes are vendor/product selection, systems integration (very different processes from systems development), systems maintenance, end-user support, and data center operations.

Since you can’t fix everything, choose the one or two processes that are broken the worst and include initiatives to fix them in your integrated plan.

How can you tell if a process is broken? Simple — your performance measures for that process don’t show an improvement trend. Don’t have process performance measures? Your integrated plan just got another entry: “Develop process performance measures.” Just remember:

• What you can’t measure, you can’t manage, but when you mismeasure, you mismanage.

• Measure what’s important, not what’s easy to measure. The two are mutually exclusive because the more important something is, the more subjective and less measurable it will be.

(Please note: Performance measures are for managing processes, not people, so focus on creating process measures, not staff performance measures. They’re two very different issues.)

Underlying all of your core processes are a few important organizational capabilities. I’m constantly surprised at how many IS organizations aren’t very good at decision-making, project management, data modeling and data management, systems library management, and training, but it’s true nonetheless. If you’re deficient in these capabilities, you won’t be able to improve your processes, so make sure your integrated plan addresses any problems with them.

Not everything is a process, no process fits every situation, and no process is good enough to survive employees who neither understand it nor care about its success. Process isn’t everything.

Saddling your employees with bad processes is like handicapping them with handcuffs and leg-irons, though, so make sure you give them processes that help them get their work done instead.