While the first part of the value cases of Process Mining focused on “Process discovery” and “Process improvement”, this second part further builds on process improvement; it also describes value cases based on “Conformance checking” and ends with cases concerning “Operational support”, more particularly predictive analytics.
Process improvement cases (continued)
One of the consequences of process variability is the formation of so called bottlenecks. When a bottleneck is structural, i.e. when it most often occurs at the same process step, it is rather ‘easy’ to investigate the cause(s). Note that this is one of the main principles of the “Theory of Constraints (ToC)” : discovering and solving the most impacting constraint in a process one by one. An additional challenge however is that a bottleneck is not always structural, i.e. at the same place within a business process.
Below diagrams – issued from Disco animated graphs – illustrate bottlenecks occuring at different times for a same process : the blue rectangle in Fig.1 between activities 3 and 4 at ‘time A’ versus the red rectangle in Fig.2 between activities 2 and 3 at ‘time B’. The advantage of Process Mining is that you can more easily discover all the bottlenecks, at anytime. Of course, this eases the Process professional to drill-down to specific underlying cases, so to assess the causes of these less structural bottlenecks.
Fig.1: Bottleneck (blue rectangle) Fig.2: Bottleneck (red) + loop (green ellips)
Such diagrams obtained by Process Mining – whether animated or not – also enable process owners aiming at “right the first time” to discover unknown loops as indicated by the green ellips in Fig.2. Indeed, such loops often go along with more expensive rework which decreases operational value (in Lean management rework is considered as the type of waste called “Defects”). In more extreme situations, process mining could even bring hidden factories to light.
Bottlenecks often occur when there is an over-utilisation of a resource; e.g. when its capacity requirement is higher than its actual capacity. However, under-utilisation of resources is also very important to know, particularly when it regularly occurs. Structural overcapacity is an inefficiency that must be managed as well, indeed. The dotted chart here above – obtained from ProM – illustrates how one can see which process activities, for which cases, a specific resource has executed: a row – like the one indicated by the green rectangle filled with yellow on its left – enables to assess whether this resource is either over-utilised or rater under-utilised, and at what moments. When hovering with the mouse, it even provides details (in the blue rectangle), like which activities of what cases it carried out, and at what time.
On top of this, a tool like ProM also provides algorithms to analyse interactions between resources, like illustrated by the so called “social network” graph here at the right. From this example, it is clear that the 3 resources indicated by a blue rectangle seem to be important ‘hubs’ in the process, given the numerous arrows pointing to those.
Compliance / conformance checking cases
Like described in previous blog, conformance checking aims at assessing whether the executed process instances conform to the process as defined. Here are the main value cases for this category:
Unneeded to tell you how important compliance is for legal and governance aspects. Think about SOX (Sarbanes-Oxley Act), US-GAAP reporting or IFRS reporting, etc. Process Mining can help you to spot process instances which did not follow the business process ‘as defined’, say according to SOX, GAAP, IFRS or any other standard or business specific governance rules.
However, nonconformities are also a key concept for Quality Management. Though a nonconformity often relates to a deviation from a specification, a standard, or an expectation for a product or service (i.e. the output of a process), its cause often goes along with a nonconformity in the process itself.
Hence, discovering non conformances in a process is highly valuable. The graph at the left illustrates how the ProM plug-in “Replay a log on Petri Net for Conformance analysis” helps to identify deviations. Attention should be paid to the red rectangles, indicating the transitions (say groups of activities) where deviations occur.
Zooming on the Statistics of such a specific transition even allows you to assess how frequently a process is subject to non conformance. The so called fitness (either move-log fitness or move-model fitness) all refer to the degree of conformance.
Monitoring business transformation
Imagine that your company is undergoing a business transformation, for instance initiated by a new strategy due to important market changes. As explained in the blog of 5 February 2015, this kind of transformation will most often lead to changes in business processes.
Wouldn’t it be wise to monitor whether the newly defined processes – following this transformation – are well respected? Wouldn’t this increase the probabiltiy for the business transformation to be successful?
Previous value cases were mainly based on so called “post mortem data”, which means that the process mining algorithms use data (event logs) to analyse past situations, e.g. process instances that have been executed. However, just like any forecasting technique, process mining can also be used to predict future process behavior. When using “pre mortem data” i.e. current data in a process “at runtime” (= before the process is fully executed), you may be able to anticipate – say avoid – undesirable situations. Here are some examples illustrating the predictive use of Process Mining:
Imagine that following an agreement with customers, the maximum throughput time of a business process is 4 days, meaning it must be executed within 4 days after its start, thus within 4 days after the customer ordered. From (past) event logs, you evaluated that executing the critical part of a process usually takes 2 days. This part is clearly the “critical path” of the process. As soon as the critical path takes more than 2 days, it is obvious that the (4 days) deadline is in danger. Hence, foreseeing a timer that warns you as soon as the critical path takes a bit more than 2 days in a running process may help you to avoid missing the deadline.
Predicting Costs of a process instance
When – based on past event logs -, you discovered that specific conditions affect the cost of your business process, you could estimate the cost of the process instance at runtime, and you may even try to influence the process to choose the variant at the lowest cost.
Example: assume that for a transport process, you discovered that for a specific destination, transport cost is much higher when there are considerable traffic jams. Then you may decide not to run this process instance when informed, e.g. through RDS, that there is a traffic jam between your company and this destination.
Predicting quality of the process output
Similarly, you may discover – thanks to Process Mining – that specific parameters impact the quality of the process output; in which case, you could redesign the process depending of the fullfilment of these parameters.
Example: imagine that you discovered – through root-cause analysis on past event logs – that the rate of rejections of finished ‘product A’ is much higher when the Relative Humidity of the air is above 90%, in a manufacturing building without air conditioning. While the quality of all other products is not impacted by humidity. Wouldn’t it be wise to be warned as soon as the humidity reaches 90% (or even 85%), so to make sure that you do not manufacture ‘product A’ at that moment, as this eventually leads to waste due to rejections? You may then decide to manufacture any other product instead.
I hope these many value cases of process mining have shed light on how this technology may help you to further improve your business processes and your overall organisation.
Next blog will describe how process mining can contribute as well to organisations applying Systems Thinking principles.
Any question or thoughts on this blog? Please write those down in below Comment box, and get free documentation in return on how to obtain BPMN diagrams from Event Logs thanks to Process Mining.
P.S.: Please share this information with your Facebookfriends and fans, LinkedIn contacts, Twitter followers and Google+ circles, through the share buttons below. Thank you!