Process Mining Hype

This challenge is also taken up by the analytical approach of process mining. The basic idea is impressive. From the automated collection of logged event data, statements are derived for an automated analysis of business processes. Free of any conceptual and subjective transfiguration - as they actually took place. A lifelike image of IT system processes at the push of a button.

This method sounds simple, purposeful and gives hope of being faster and more efficient than any manual or interview-based inventory of operational workflows (as-is analysis). Not least thanks to this idea, process mining has found a firm place in the analysis of business application systems. This is because process mining helps to discover, reconstruct and monitor real processes close to reality.

Historically, this method is not a new approach, but looks back on many years of development. Opinions differ as to exactly when the idea of process mining emerged, whether it goes back to the 1990s or emerged in the 2000s from research projects that pursued the objective of developing new meta-models and ontologies for process flows from workflow-based analyses.

There are many reasons for conducting such an investigation: Existing processes are increasingly inefficient and cost-intensive due to creeping changes in requirements over time; process flows and responsibilities are unclear, especially against the backdrop of distributed organizational units; heterogeneity exists within the enterprise group, perhaps due to growth; company-specific or vendor best practices are not realized; a low degree of standardization exists, possibly due to unstructured and non-strategic business requirements; the degree of automation to reduce process costs is too low; or process costs are not distributed according to their benefits. This is more than enough motivation to strive for complete transparency of business processes.

Event logs and visualization

But does process mining really do justice to these tasks and the current hype? How meaningful and reliable are the results of process mining? And is process mining the same as process mining?

To answer these questions, a basic understanding of how it works is required. Event logs, which log the process usage by employees, form the information basis. This can be used to generate visualizations that provide a correct and up-to-date overview. This is exactly the strength of process mining, because by using actual business transaction data, an objective picture of the process is created. The extraction of this information is usually done either by a full deduction or the direct connection to the system under investigation.

In any case, complete access to all individual document data is required in order to ensure completeness of the process chains and to be able to process the initial data preparation in an analytically meaningful way. The focus is on the transaction data for which a context is subsequently created (filtered), for example, by means of master data such as customers, materials or organizational units. 

An automatic, generic structuring and classification of the data as activities and paths seem rather cumbersome and the question arises whether the consideration of event logs does not disregard meaningful aspects of an SAP system in particular, such as the parameters of the system configuration. This is exactly where classical process mining reaches its limits, a misunderstanding that will be examined together with some others in the following.

Six misconceptions

Misunderstanding 1: Process mining delivers fast problem diagnoses and bottleneck identifications. The basic assumption that weak points in day-to-day business are unknown and must first be identified has been disproved in many projects. An SAP system provides the individual employee with a solid overview of the inconsistencies in his or her work area, even if not of structural causes and the full extent of their integration effects. Accordingly, statements made initially often belong to the realm of trivial observations or deliver false positives. Costly investigation iterations are the result. This can be remedied by a comprehensive analysis model that is based on a broad set of key figures and is available right from the start. However, this model must be compatible with both the specifics of the company and the target concepts stored in the customizing of the software. Otherwise, there is a risk of cost-intensive efforts for subject matter experts and subsequent adjustments of the analytical target parameters.

Misunderstanding 2: Process mining provides a quick overview. Many process mining providers deliver the technical capabilities for analytics, but include only simple templates without any focus on the specific customer situation or the differentiated capabilities of the software. This design is also often shifted to the area of paid services. It is advisable to check here exactly how extensive the available analytical content is and which competencies come into play at what cost with corresponding service offerings.

Misunderstanding 3: Process mining helps to assess compliance, risks or otherwise. Of course, process mining can also help to assess opportunities and risks of business activities as well as to investigate compliance. Finally, the analysis accesses the entirety of process information. But the application of test logic requires sound data transformation, analysis, preparation and evaluation, as well as consideration of the individual characteristics of what company-specific risks or requirements actually are. Accordingly, these statements are also strongly dependent on the underlying reference content or supplementary services. Sometimes promises of the software manufacturers leave the actual of this approach.

Misunderstanding 4: Process mining allows process data to be consolidated across multiple systems. The wishful thinking of a central ERP system is regularly not given, especially in large companies or corporate groups. In practice, there are distinctive system landscapes with distributed processes. For their analysis, classical process mining is rather unsuitable, because the reconciliation of cross-system process chains is associated with high effort, especially in the case of variant-rich process logics and heterogeneous technical systems. A direct comparison option at the level of system technology and organizational units could help here and also offer approaches for consolidation and harmonization projects, but this is available in very few cases.

Misconception 5: Process mining provides an end-to-end view for companies. The desire for a complete view of all business processes and all work steps involved in an enterprise process is understandable, but requires multiple levels of process analysis. Especially for documentation purposes, test efforts and change discussions, such scalability is useful and desirable, especially for integrated and heterogeneous areas, such as links between logistics and finance. However, building such a representation is very complex and requires some cross-disciplinary knowledge. For this reason, process mining rarely uses end-to-end models, but rather fragmented sub-processes. 

Misunderstanding 6: Process mining supports the automation of processes. Automation in the sense of the elimination of functions or process steps that are no longer executed by dialog users but by the software itself can certainly be measured by process mining approaches. However, the implementation and improvement of these processes ultimately fall into the realm of consultant knowledge, all the more so because this requires clear coordination of the involved company departments and their internal work processes.

Reference and target models

In summary, the suitability of specific process mining approaches should be assessed on the basis of content-related aspects. First, it is the analytical reference models for structuring the process data and deriving business statements that set the quality standard. This includes, in particular, the availability of software-based target models as well as correct and comprehensive performance measurement systems that can look behind the scenes.

In addition, it is the type of extraction, transformation and preparation of the technical database that determines the quality and correctness of the evaluation options. A fully comprehensive database does not always meet the cost-benefit requirements and requires some work and expert knowledge to develop the desired informative value. Enriching the event log information with change histories, benchmarking or configuration information can be helpful here and significantly accelerate the gain of knowledge. 

Nevertheless, the necessities to find good consultants and internal resources remain and elaborate modeling efforts for individualization have to be carried out. As a tip for the initial deployment, it is advisable to focus process mining projects. First obtain an overview of process usage and examine the existing content in greater depth where there are highly individualized processes.