Process analysis and optimization is a costly endeavour. Besides the vast amount of resources it consumes, today’s – to a large extend manual – analysis often relies on subjective views. The realized benefits are often unclear and not quantified. Yet, the vast amount of unused process data could provide analysts with qualified information regarding optimization potential within seconds.
While LANA’s automated actual-target-analysis is already eliminating a large part of manual effort, our vision and the technically-possible goes way beyond that. Therefore, we aim to further push forward the automation potential in process analytics. Our latest achievement in this regard is the detection of root causes in a process by employing classification algorithms from the data mining domain. With LANA we offer a solution to instantly detect deviations of a process, but so far the tracking of the root causes still had to be done manually. With the new automated root-cause analysis (ARCA) for LANA this changed. An automated root cause analysis is largely expanding the capabilities of the current state-of-the-art in process mining applications with far reaching implications.
Retrieving Data for the Root Cause Analysis
The main target of a root cause analysis is to spot the factors that caused detected and observed problems and provide information on their nature. Here, especially the magnitude, location, timing, and context of caused problems are of importance. Typically, this information can be extracted from the process execution data recorded by the supporting IT system. With respect to data extraction, there are multiple options that are already successfully applied in practice:
- Per-analysis extraction using simple database queries,
- regular automated extractions using so-called ETL tools known from business intelligence, or
- direct integration of the extraction logic into the analysis tool using existing interfaces.
Yet, the extraction of often highly customized IT Systems, such as SAP, requires some expertise and thus, initial implementation projects for each customer. In contrast, the integration of highly standardized cloud services enables data import by mouse click, once the interface is provided.
The figure above shows how the process execution data shall be leveraged. First, the already existing actual-target analysis is used to automatically derive the problem that needs to be analysed. In this step, the process data is classified and subdivided into cases in that the problem occurs and cases where it doesn’t.
In the second step, the classified process data is used to identify correlations of provided factors and the occurrence of the problem. To this end, classification algorithms from the data mining domain are leveraged. This classification results in a list of factors that strongly correlate with the problem and can therefore be seen as potential root causes. These root causes can be found in used material or equipment, people, process execution, management and environment. Every process execution is related to different attribute data, such as the affected product or country, production site or involved machines. For example, the root cause for the problem of long production cycles might be strongly correlated with the use of a particular machine (equipment) that regularly overheats (environment).
The results of the root cause analysis are the input for improvement measures and pinpoint what behaviours, actions, data, resources or similar related aspects have to be altered to avoid the reoccurrence of such undesired behaviour in the future and derive best practices and assert faultless execution.
Automation enables process mining for everyone
The automated root cause analysis is further eliminating manual analysis in process mining tools and instead provides the analyst with process violations and their root causes in an automated fashion. With further automation, process mining will become attractive for a user group that the method has not yet been accessible to.
The user doesn’t need any specialized analysis capabilities himself and can completely focus on improving his business processes. With a tool like LANA, the advantages of process mining can be made available to a much larger target group as less knowledge and even fewer resources are needed.
For a company that does not have the knowledge or resources for pursuing active process analysis, LANA provides a huge potential as the automated process analysis simply presents a list to the user with prioritized issues, their location in the process (activity level) and their root causes. This list enables to efficiently design fact-based optimization strategies and assuring high process quality in terms of timeliness, efficacy, compliance and resourcefulness.
In companies that are actively managing their processes, the effort for regular process analysis is dramatically cut down and resources can be located to actually improving the processes, while monitoring the changes over time.