In his excellent posting, Keith Swenson makesmany good points. He points to the range of interpretations of BPM, and particularly highlights the issues associated with its interpretation by software engineers as just another piece of hype on the road to good programs. But I think there is another, perhaps more important strand that is buried in there. As Ketih points out, BPM is about the Management of Business Processes.
As we all know, everyone’s interpretation of the term Business Process is different. In my training (whether that be BPMN, or higher level training on BPM Methods), it is one of the first things I get people to write down (inside the first 5 mins), and not surprise, every definition is entirely different. And when those people are senior managers in a ciompany, their interpretation of the term is invariably what I call “Process as Purpose“. The point is that they see Processes as being more about the purpose than the constraint implied by sequencing of steps. They are there (at the training), because they see the importance of “Managing” their processes. Indeed that concept (Managing Business Processes) as central to the success of their companies.
[[ I am still down in Brazil, and I am really struck by the process sophistication of the people I am meeting. They all get it. I was more than surprised to find two "Business Owners" (people who own significant businesses), giving up 3 days of their time to come on a course around how to structure and run BPM programs. They were cherry picking from the broad range of techniques we covered, but ask yourself whether you could imagine the CEO or COO deciding that they should attend a public training course. That's what I am getting at about the sophistication of the Brazilian business climate. ]]
Coming back to Keith’s post, he describes a spectrum of BPM interpretation – from pure Software Engineering (where the SW Eng tries to reflect the needs of the business person’s Business Process); through Business Processes being modeled by a business person, then tossed over the fence to a Software Engineer to finish them off; to the Business Process as modeled by the business person, then being directly executed (what he called ”Pure BPM”). I am not quite sure I agree with the Pure BPM bit, but I do know what he is pointing to … where the processes of the firm are driven by models (without translation to some intermediate executable format (like say BPEL).
One of the comments on Keith’s post points to the challenges of getting business people to model their own processes and make the resulting collection of stuff useful. He described the usually resulting mess as an “expensive disaster”. And the reason for this is that business people dont usually have the sophistication to understand their business problem as a set of inter-releated processes that between them deliver on the “Process as Purpose” concept I referred to earlier.
Invariably, process modelers (whether IT or business) tend to see a process problem as a single process. They interpret the high level Purpose as a single implementation process (which invariably it is not). They make all sorts of mistakes such as mixing up the ”Handle an Instance” with the “Manage the Flow of instances”; they switch from batch mode to handling a single instance; they dont think about the interfaces between processes (handing information from one to another), etc. What they do is try and connect up everything that sounds like an Activity into one convoluted process.
Now software engineers are usually more adept at the necessary abstract thinking, but that doesn’t mean to say that business people cannot wrap their pretty heads around the notions. It is merely a reflection of the fact that they have not had adequate training. What is missing (across this entire industry) is better learning around “Process Architecture” – what “chunks” do you need and why. Poor chunking leads to unnecessary complexity (and even “expensive disasters”).
We are still stuck with decomposition as the prevailing mind set – where sub-processes are always contained within the parent. SOA concepts seek to get around this, but there is also a higher level “Business Services Oriented Architecture”. Processes should not be regarded as some sort of static hierarchy, they are more accurately regarded as a network of interacting instances. Think more jigsaw puzzle than org chart.
When I gave a “Power Breakfast” at the last Gartner BPM Summit on BPM and Process Architecture I had a packed room (it was starting at 7:30 in the morning so these people were keen). I described a set of methods that you could use to go from “What Business Are You In” to what “Processes Do You Need” right down to the SOA components if that is what you wanted to do (I would recommend looking at a BPM Suite first rather than going straight to the SOA software engineers paradise). I only saw one person out of the 90 or so get up and leave, and nearly everyone else gave me their card at the end of it. The room really was comprised of mostly Enterprise Architecture folks from the IT community, all of whom struggle with this transition.
Switching tacks – the vendors BPM Suites are unconciously making this architecture problem worse. With only a few exceptions (Pega, Itensil, BizAgi … I am sure there are others in this category too, these are the ones that spring to mind), vendors interpret the business process problem as being entirely seperate from the data and artifacts associated with the process (business people see them as intertwined). They regard the process relevant data as a set of named value pairs … the information required by Process A is declared on Process A, and must be recreated on Process B and then mapped from one to the other (if the processes need to communicate with each other). That means that there is an extra (unnecessary) layer of complexity for business people trying to reflect their business problem. Moreover, if you change one process, then you need to refactor all the interfaces. This is “software engineering” oriented thinking.
The other approach is to define your data structures (perhaps as an “Entity” defined as an exstensible set of XML artifacts) and then describe the views on those artifacts at the level of the data structure. Then it is merely a matter of associating your processes with the Data Entity, and all the different views become available. Process interfaces become an order of magnitude more accessible (to the business user), you can use any number of processes to support a single case of work, and again it helps move away from the software engineering mindset we find in so many BPM tools (which were often created to solve the problem of Enterprise Applicatin Integration … hence their association with Software Engineering).