Process mining has now reached almost all DAX companies and many medium-sized companies also use this analysis method to get more transparency in their processes. However, technology is currently developing rapidly. The expansion goes in the direction of advanced process mining with additional functions and new possibilities.
With new facets such as built-in ETL tools, predictive process analytics and task mining, it is not exactly easier to correctly assess the potential of process mining for your own company. When starting, it is important to pay attention to the pitfalls and to put the projects on the right track from the start. This particularly concerns data engineering and questions surrounding data sovereignty.
Process mining is – in contrast to classic business intelligence – a representation of data that is characterized by process flow diagrams. The process data reconstructed in this way is graphically displayed in a process mining tool, in particular using process flow diagrams, which can be filtered and analyzed for their waypoints and flow times.
Technically speaking, process mining is essentially based on a graph analysis, whereby process activities, nodes and the chains of processes represent the edges of the graph, with parameters such as the frequency of occurrence or the elapsed time between the activities. In a broader sense, however, process mining also includes the entire analysis of processes, including at the KPI level and pattern recognition, for example fraud patterns.
One should not lose sight of the fact that process mining does not purely model specific process models, but rather works them out of the operative IT systems. For example, if you don’t find your Kanban process in the technically validated process representation, you actually don’t have the implementation of this concept of process organization. A common feature of many tools is therefore the comparison of the reconstructed actual processes with theoretical process models (target processes). Process mining can also be used to analyze existing processes and identify their worst and best practices via analysis of all process variants. If necessary, a new target process can be developed from this.
Well-known providers of such tools include Celonis, Signavio, UiPath, PAF, MEHRWERK, Fluxicon and Lana Labs. The tools are very different from each other in terms of functionality, cloud and enterprise capability. Process mining is a catchphrase that a number of BI tool providers are also committed to. Process mining is an analysis methodology that involves the reconstruction of processes from log data and other data tracks in IT systems.
Even if the word reference is obvious – process mining has hardly anything to do with data mining. The latter forms a collection of methods for mathematical algorithms for unsupervised machine learning, for example DBSCAN, K-Means or PCA. It is interesting, however, that the mathematical data mining methods and predictive analytics are now also being used in process mining, but more on that later.
Currently, process mining can be described as a process that requires at least one data engineer to combine data into a protocol using SQL or a programming language such as Python. It is all about data engineering. Data series are linked as process activities (events) per time stamp (timestamps) using identified process numbers (case IDs). The processes can then be analyzed on the basis of this preliminary work.
The tool providers for process mining originally only supplied the tip of the iceberg. Even if they have long offered interfaces and automated data processing for certain IT systems and classic processes, primarily in the area of working capital management for standard ERP solutions, most process mining projects are still very busy with the raw data Level connected.
Process mining and business intelligence have so far mostly been discussed and applied separately. Business Intelligence (BI) deals in particular with the structured provision of at least daily reports or dashboards that provide quantitative information about the current company situation. Business Intelligence traditionally works retrospectively or can report on the current situation in almost real time, as is the case with process mining. Process mining is a sub-discipline of BI that does not focus on general or finance-oriented KPIs, but on process-specific KPIs.
Some providers have been working for a long time to combine process mining with business intelligence and to offer both disciplines integrated in one solution. The external appearance of business intelligence is characterized by dashboards that use BI tools to display information with tables, bar charts and histograms. They show data in absolute and relative values as well as their distributions.
This contrasts with providers of classic BI tools such as Qlik, MicroStrategy, Tableau and Microsoft. In some cases, the process mining providers enter into close cooperations.
With its product “PAFnow”, for example, PAF (Process Analytics Factory) relies entirely on the Microsoft stack and PowerBI, which is supplemented by a PAFnow plugin to become a flexible process mining tool. This plugin approach has the advantage that the user can directly use the widely used BI tool PowerBI, which also has a large user community and many interfaces, for process mining.
Mehrwerk follows a similar path and offers process mining as an extension to QlikSense, also a common BI tool from the software company QlikTech. According to Mehrwerk, this solution focuses on the synergy of BI and process mining as well as on simple data preparation and data governance via Qlik.
Lana Labs relies on open interfaces for integration to enable seamless integration into individual IT landscapes. The software makes it possible to integrate process mining results into the existing BI dashboards (Qlik, PowerBI, Tableau, etc.) through its interfaces.
- Lars Schwabe (Associate Director at Lufthansa Industry Solutions
“The success rate of predictive analytics projects has increased since the companies have finally done the necessary preparatory work, for example the creation of modern data architectures. In addition, both the staff have become more knowledgeable and the tools have improved. “
- Daniel Eiduzzis (Solution Architect Analytics at Datavard)
“Technically, companies have to open up and should not slavishly commit to a manufacturer. Today, it is more a matter of identifying the ideal instrument, depending on the respective use case, with which the questions are served as best as possible. Therefore, a best-of-breed approach can make sense here. “
- Jan Henrik Fischer (Head of Business Intelligence & Big Data at Seven Principles)
“With methods of predictive analytics and the increasing digitization in parallel, we will understand processes better. Without exception, this will affect all areas of a company. The greatest potential lies in the optimization of customer processes. With a deeper understanding of their needs, we will be able to serve customers more efficiently and better, and increase their loyalty. ”
- Vladislav Malicevic (Vice President Development & Support at Jedox)
“Many companies have been experimenting with predictive analytics for a long time. So far, there has often been a lack of concrete use cases with a clear added value, the so-called business case. But the next phase in the technology lifecycle has already begun, and companies are no longer just conducting innovation-driven experiments. They increasingly link predictive analytics and AI projects to a clearly defined added value for certain specialist areas or business processes, including the expected results and the possible effects on previous processes. “
In the future, BI systems will also be supplemented by predictive analyzes. These include, for example, forecasting models for sales, warranty claims or purchasing requirements. Process mining is already supplemented with data science methods to predict throughput or waiting times, for example, with predictive analytics (supervised machine learning) and to detect anomalies in process chains, for example, for process optimization or also with data mining (unsupervised machine learning) can play a role in fraud detection. In the meantime, Celonis has integrated the first functions of machine learning with its operational apps as well as UiPath, Lana Labs, Mehrwerk and PAFnow.
Overall, this is just the beginning of continuous development. In addition, experienced data scientists can use the event log to start manually and realize their own ideas even more individually.
But the other tool providers are also working on the integration of machine learning. For example, Lana Labs’ R / Python integration enables users to access a wide range of AI algorithms in addition to their own machine learning functions and to expand process mining with their own developments. Other providers follow suit and the solutions based on PowerBi and QlikSense also benefit from the R / Python integration of this BI software.