Most business process modeling processes entail (at least in the beginning) redoing existing processes rather then creating completely new processes from scratch. I have a computer science\HCI (Human Computer Interface) background so I think of the first step in managing a process (modelling the existing process) as the requirements gathering phase. For human processes, it is very similar to requirements gathering for HCI – the best way to do this is to go out and see how people get the work done today and use that as the basis for process model (the process you’ll find is probably sub-optimal, but it does work or the business would be failing). Elsie Olding has a good post on this in Jum Sinur’s blog.

Once you have a model – how do you know you got it right? How do you know you have covered all the basis, and that you have really captured the process – not what people imagine the process should be? This brings me back to the notion of iterative (or agile) design methodologies. There is the need for some type of user validation step (in HCI we would use low and high fidelity prototypes), that as far as I can tell is missing in most BPM projects (beyond having the users look at the models themselves- which most people find difficult to follow).  One of the problems with no validation step is that  people are pretty good about the applications they use, but they are a lot fuzzier when it comes to the handoffs, or workflow – especially for exceptions. Without the ability to actually validate what they say, it is hard to know if they missed something or left something out. It is almost like trying to recreate an application by only using what the users tell you – you are sure to miss some important features.

One way to track handoffs would be to create a quick implementation of the process frameworks using ActionBase – creating a few ActionMail types and giving the process participants the ActionMail plugin to use instead of regular email. If you want to also track application usage you could have them install an application usage tracker (like Wakoopa) and then correlate the two information sources (ActionBase and Wakoopa). Doing it this way (iterating on the as-is model based on user info and actual fact) would ensure that you really captured the as-is process -and validated your process model.

After the model is validated, only then would I move on to an optimization step – and again I think iterative, agile methods are the best. Being able to iteratively design process modifications is going to be key in making human (or unstructured) processes manageable. There needs to be a lightweight method (under human supervision) to tweak the process and test the results before you commit any optimizations.