Is all History Bunk?
I think he has some points, but he is also missing quite a few.
Let’s start with the feasibility. He correctly identifies two major problems that occur in practice:
Data can be incomplete; different systems contain different kinds of data.
The data may not reflect the actual process.
The first problem relates to the tracking of a process across different (parts of the) IT systems. For example, one needs to link the purchase order numbers with the relevant invoice IDs to look at the whole purchase-to-pay process. This can be a challenge.
However, while he argues that:
What is kept in one system is missing in the second, and vice versa. Eventually you’ll find yourself on the smallest common denominator.
The minimal process mining requirements are actually quite basic: You need to be able to identify process instances (case IDs), activities (process steps), and ideally some timestamps. If it is really not possible to link the, e.g., purchase and pay processes together, then it can be still valuable to analyze these sub processes in isolation.
The second problem is more severe than missing information. Missing data is obvious, but how do you know that the data reflects the truth? He gives some nice examples of how systems can be misused and probably many people can tell similar stories.
But is this a reason to not use the data at all, without even looking at how good they are? In fact, the data that are collected in organizations are constantly getting better—it’s a byproduct of the increased IT support but also because it is understood how relevant they are for decision making. And: The use of process mining tools does not imply that it is forbidden to talk to people and validate the results.
As with everything, one needs to decide on a case per case basis what is the trade-off between effort and benefit.
So, what are the benefits? And here is where I disagree with Ed the most. He basically states that looking at the past has not much use.
But Process Mining only provides information about how things were and not how they will be. And the latter is precisely what management is most interested in.
There are situations in which the “as is” process may not be relevant: If the process is not important, if it is running smoothly, or if one deliberately wants to redesign it from scratch without being influenced by how things are done right now.
However, in most situations getting a clear picture of the current process reality is both necessary and extremely difficult. It is necessary to be in control and to be able to improve. It is difficult because processes are in the heads of many people, distributed over different departments and IT systems. While one could easily walk along a manufacturing process, particularly service processes are inherently invisible.
Process mining can make a crucial contribution by providing a fact-based rather than an interpreted picture of the current process.
Here are a few benefits:
Less time is needed from the process experts in interviews or workshops (in which they are away from their jobs) to understand the “as-is” process. The process mining work upfront leads to less intrusion in the actual business.
Discussions that are based on opinions around possible root causes of problems can be avoided or resolved. The fact-based analysis provides an objective reference point.
Not just the sunny day scenarios but also the exceptions become visible. The data analysis reveals all the variations of the process, and often the 20% that are “non-standard” cause 80% of the costs.
Unlike with standard query tools you don’t need to know upfront what you are looking for. By showing what is actually happening, unexpected insights can be obtained that one would have never thought of checking explicitly.
Where does this leave us?
Ed calls process mining a hype, which surprises me. I wonder what he has heard, and from whom?
Process mining by itself is not a silver bullet. It is not a methodology, and it does not seek to replace other techniques such as data mining or business intelligence.
Instead, it is a great tool with real, tangible benefits that can be combined with data mining, embedded in the lean six sigma methods, and used on top of BI and data warehousing technology.
Let’s all be careful, while spreading the word, not to overpromise. It would be a pity to discount process mining as a hype before organizations have started to get a chance to reap its benefits.