This month I have been preparing for a presentation at BTOES Process Mining Live. During research for the talk, I was struck by how it seems history is about to repeat itself, and it looks another religion war may be about to break out.
Years ago, in the early days of BPM, the war was between modeling (mainly BPMN) and the low/no technology approaches like Six Sigma and Lean. Over the years, the two seem to have found a more peaceful coexistence, with both sides agreeing that there is value in process modeling, but only if you leverage the wisdom of the crowd, and use process modeling as a do with not a do to approach. Many in the modeling community are also becoming more relaxed, accepting that every process does not have to be modeled, and sometimes s simple sketch is enough.
Process mining, which advocates agree, is a technique that uncovers digital traces to produce models is the new kid on the block. Reconstructing the links of chains between these traces from system logs provides highly useful insights into the different paths users take through a system to undertake work. Suggested benefits include greater ability to reduce variation, ensure greater compliance, understanding bottlenecks, and identifying potential areas for improvement. Interestingly enough, though for a technology that will soon approach 20 years old, there are still relatively few case studies available.
The religion war starts when those mining advocates transpose language. Instead of talking about digital traces and transaction chains, which are what is discovered, they talk about them as being process models. Further, many proponents of process mining suggest that these machine captured and visualized views actually replace and do away with the need for process modeling tools. They argue that process modeling tools don't represent reality and that only mined models are real. The belief is that process modeling is about theory, and process mining is about facts. To some extent, that may be correct in some cases. In others, the facts don't always support the changes required or the story to be told. Maybe sometimes they are what we might call alternative facts.
I suggest this is a dangerous path to go down. Process mining is still in its infancy, with many tools still needing to add depth before generating the revenues vendors aspire too. It does not make sense to me for these niche vendors to pick fights with modeling tools and the broader process community. Far better for process mining to carve out and deliver ways to supplement what process professionals and teams currently do and the tools they use. There are plenty of gaps in the offerings of modeling tools, things like reporting, dash-boarding, and monitoring spring to mind, which are fertile areas for growth and are well suited to applying process mining technology. Indeed some vendors already recognize this and have focused their efforts on connecting process mining with content, enabling far greater value when using the mined data, while others are broadening their offerings to mine tasks and journeys.
Conversely, modeling tool vendors that eschew the automated discovery that process mining adds miss a trick. The ability to automatically capture and analyze some aspects of the process can accelerate many process improvement initiatives. To then be able to connect processes together, undertake resource analysis, and conduct simulations on multiple scenarios is the added value modeling vendors can bring.
There are many process mining demos I have seen where the bottleneck analysis improvements shown are flawed. In part, this is because process mining tools don't tend to understand enough about resource availability, or what manual work takes place along the way. Not all work is digital, and many processes span systems, span manual and automated tasks so often, the need is for models that can combine both. Sure, there are occasions when digital traces are enough, just as there are times when whiteboarding a process is enough. The process teams that will win in the long term, now as ever, will be those whose toolbox has all the tools they need. You can't rely on a multi-tool where the vendor suggests using a wrench as a hammer or tells you the hammer can replace your screwdriver.
The world has started waking up to the importance of process as a business differentiator. The last thing we need is for businesses to get caught in another religion war that results in tooling being placed in the “too difficult pile.”
So, let's not confuse digital transaction chains with process models; let's not pretend it's a binary choice of one or the other. Instead, let's focus on helping people better understand when best to apply the appropriate tools.
There is lots of fertile ground where both camps can do more to deliver better business value while swimming in their own lane, or maybe an empty lane beside them.
One example of swimming in the same lane is the use case of Digital Transformation, both modeling and mining vendors seem to push that as a business problem they address. I can certainly see that models help to understand what can be done differently and which parts of the business can be digitized, and which parts can be transformed through digital technology. Process mining vendors are also pushing their technology as key to this use case. Yet, if what they are discovering are digital traces, then the business must indeed have already transformed? In this case, their value is not so much in the Digital Transformation itself, but in the digitally transformed organization's monitoring and management. Adding Process Intelligence, as I think of it, is a great way to monitor and maintain the gains and to move from transformation to continuous improvement. For me, this is an example of an empty, and adjacent lane with apparent value, rather than a competing lane. Perhaps, an example of a cooperating scenario.
Taking another use case as an example, most modeling tool vendors see the opportunity of S4/Hana migrations for SAP customers are fertile ground—huge projects with lots of spend, and thus opportunity. To date, I have not seen many process mining vendors playing to this use case. Yet here is an example where maybe modeling is not as appropriate as mining. For many organizations, this migration is less about transformation and more about migration. With considerable investments in SAP customizations over many years, many don't know what they might need to carry forward and what they can leave behind. In my mind, always a dangerous place; this is something that a process mining approach is much better suited to than a modeling approach. By mining the transaction chains, we can quickly identify which customizations are used, how frequently they are used, and the impact of not having them.
Speak Your Mind