This is a common challenge once a company has built enough “stuff”. The burden of maintenance and the difficulties in refactoring makes one cry out for “Simplicity, now!’. A common perception of reducing complexity is to “dumb things down”, this is a reaction to the tension of it being difficult to untangle things. It’s often a good idea not to act on this tension, but look a bit further at the issue at hand.

Before even starting to consider doing anything, it is worth while to take a step back and analyse what exactly is complexity. People working together sometimes create established truths and create a shared understanding of how things are. Not based on research or analysis. There are multiple forces at play when a team starts talking about “too much complexity”. It can be a failure to deliver on their promises leading to a sense of failure and a need to place blame outside themselves. In companies with a hierarchy teams often create a lore of complexity as an explanation of why things are not going as planned. In the post The hidden meaning of complexity in the context of software development I talk about how the word complexity can mean something different entirely.

How to go about changing things?

One way to reduce complexity is to have to deal with less. Are there parts of the application which hardly ever changes? Would it make sense to split out isolated behavior from the main build? When doing this you can often get the counter argument of “we don’t want micro services!”. True, splitting out into separate might reduce some complexity and focused services that hardly gets deployed are great! However you can also isolate code which hardly change into libraries you compile in, reducing the number of things to build.

The misconception of thinking that reducing complexity always means removing or reducing things. You might actually reduce complexity by embracing new paradigms and undertake larger changes in your architecture. Doing this might require you to acquire new capabilities and learn new things, however this is not the same as adding complexity. Learning and evolving is always a key part of software development and should be just as natural to do when trying to reduce complexity.

Let’s say you have monolithic API which has grown out of control with long build times and a lack of efficient horizontal scaling, one way of reducing complexity could be to make drastic changes. Perhaps you need to insert a message queue and embrace an event based architecture? This might seem counter intuitive to reducing comp_lexi_ty , as setting this up is not that simple. However I would argue that often is this planned complexity preferable to the accidental complexity of a module which has grown out of control.

The means of production

When rescuing complexity an essential task is to look at the means of production, which is software is the delivery- and runtime infrastructure. Whatever path you choose in order to reduce complexity, it has to start here. If you continuous delivery system is fragile, or non existing, that is a place to start. If your infrastructure isn’t possible to replicate without manual procedures, that’s a thing to fix.

The reason for this is simple, no matter what you’re looking at doing it’ll be a nightmare if these things aren’t running smoothly. It will greatly reduce the types of things you can choose to do and it will also increase the time it will take to dig you out of the hole you are in.