The butterfly effect

I wasn’t in the mood. I’ve been writing about shadow IT in recent columns, and discovered what follows from 21 years ago. It delivers some messages I’ve been wanting to cover without my having had to put fingers to keyboard.

Hope you like it.

– Bob

# # #

System changes are like Mexican butterflies.

For decades, we’ve been migrating functionality from applications to infrastructure. We used to choose the best application and bought whatever platform it ran on, congesting our data centers with a hodgepodge of incompatible systems. We’re smarter now … we understand that eliminating redundancy is as important in the platform layer of our technical architectures as in our data designs, so we establish technology standards and require new business applications to be compatible with them. In doing so, we’ve moved the physical interfaces, especially network and DBMS compatibility, into the technical architecture — we’ve moved functionality into the infrastructure.

With OO we create a library of reusable objects. Before OO we created subroutine libraries and COPYLIBs. These libraries are part of the infrastructure, too.

Have you installed an ERP suite? If you do, it has an API that turns it into a platform. By acting as a platform and not just an application, ERP has also moved functionality into the infrastructure.

Enterprise application integration (EAI) has the same result. It turns data and logic interfaces, which used to be part of your applications, into yet another part of the infrastructure.

Every time you move something into the infrastructure you improve maintainability, integration, consistency … all good things. Unfortunately, you also lose flexibility, because with all of these benefits comes interconnectedness. Like the butterfly of chaos theory, which supposedly messes up the weather in New England by flapping its wings in Mexico, even tiny application changes can have unexpected and significant consequences.

That’s a problem, because while we’re busily moving everything in sight into the infrastructure, increasingly sophisticated business leaders, workgroup supervisors and individual end-users spot endless opportunities for improving the business through information technology. Maybe it’s a Macintosh in marketing. Maybe it’s a tracking system for the trade-show team, a contacts database for public relations, or thought-mapping software for strategic planning.

Whatever it is, it represents a business improvement opportunity for some small community of interest in the company and another headache for you. They get the benefit, you have to maintain it, make sure it continues to operate when you upgrade hardware and operating systems, integrate it … and then it becomes part of the infrastructure, too.

It’s easier to just turn down these requests, because approving them all feels like being smothered by a swarm of Mexican Chaos Butterflies.

There is a third alternative to rejection and asphyxiation: Recognize that moving everything into the infrastructure is the equivalent of creating a centrally planned economy — in this case, an information economy. As we learned by watching the eastern bloc fall apart, when it comes to economies, central planning has its limits.

Not everything belongs in the infrastructure. Create a space outside the infrastructure for beneficial uses of information technology that just don’t fit, and don’t have to fit.

Create a multilevel support framework that establishes the ground-rules. A standalone system can be pretty much anything. If it needs to run on the network you need a traffic analysis; if it needs read-only access to existing databases you need volume estimates and adherence to security standards. If it needs to update core data … sorry, that’s a Mexican Butterfly.

For projects fitting the framework, the requestor is free to contract with an outside company — you’ll recommend one — and you’ll work with the contractor to make sure the results fit into this framework.

Don’t create the framework on your own. Develop it with key business users and ask your Systems Steering Committee to endorse it — it can’t succeed unless the rest of the business accepts responsibility and accountability for the information systems they ask for. Not everyone does. Sometimes, people asking for your help want you to turn them down. That way, they get credit for trying without the pain of change; you get the blame for being an obstacle to progress.

Turn the tables. Give them the worst thing anyone can get: Exactly what they’ve asked for.