Process Mining in Financial Institutions
A technical approach to Business Optimization and what to consider when considering it
Process Mining is often considered to be the new “holy grail” of process optimization. Our experience shows that, while it can be a powerful business analysis support, the real leverage depends on some assumptions and pre-requisites to make the most out of it. In our summary we provide a short insight into this topic.
Process Mining: What is it?
Basically, the main goal of Process Mining is the recreation of a process model from event logs. More precisely, the event logs from corporate information systems. Therefore, as a result of Process Mining, you receive a graphical representation of an as-is process as it is represented in the event logs. This generated model can then be compared to the to-be process model as starting point for process optimization. For example, an as-is process may contain redundant iteration cycles, unnecessary process steps, wrong sequence of process steps, extreme processing time values or other kinds of differences to the defined to-be process model.
How is it done?
Process Mining is normally performed with the help of a tool to design a process model and conduct analysis. First you need to load event logs in the tool and then it will assist you with the design of a process model. However, each Process Mining tool requires interactions with a human to filter data, to remove noise, analyze and correct results. Therefore, a Process Mining tool can be seen as a structured way to give you an opportunity to visualize your process model, to discover discrepancies between different process instances and to help with distinguishing bottlenecks.
When does Process Mining make sense?
While the promise of Process Mining sounds interesting and promising, there are criteria that must be met to generate value out of the exercise:
- Availability of adequate event logs – to use the event logs effectively, they must contain a minimum set of data, such as: Case ID, time stamps, activity.
- Quality of the event logs – event log data must be complete, accurate and reliable to produce a realistic process model
- End to End data – processes involving different IT application must be linked via identifiers or handled in a digital workflow to ensure an effective analysis
It is important to point out that Process Mining helps to generate real process model on operational level. The decisions made at strategic or tactical level and their rationale are not questioned or analyzed during the exercise.
Process Mining will therefore support understanding the location and quantification of a process bottleneck. It can also support the analysis of process compliance in the sense of comparing as-is processes with to-be processes. Within the broader context of process optimization and automation, Process Mining is a powerful tool to get started with analyses and to monitor effects of implemented measures. It will support visualization and discussions. Nevertheless, the actual definition of optimization measures, still must be done by a process management team.
What are the main benefits of Process Mining?
- It helps to streamline processes and offers transparent quantitative process data as basis for definition of optimization measures
- With the produced data and visualization, it is much easier to discover and eliminates non-value-added tasks and activities within as-is processes
- The same is true for visualization of unnecessary iteration cycles, which can result in quick efficiency gains with relatively small measures
- Due to the usage of real world data, performance bottlenecks can be identified and investigated rather quickly compared to traditional process analysis methods (interviews, observation, etc.)
What are the main challenges to be addressed when dealing with Process Mining?
- Event log data doesn’t include all the data of an investigated process. As event logs focus on daily activities they mainly consist of operational data whereas optimization potential can reside in other data of the process.
- Data quality is key: completeness, accuracy, timeliness, reliability has to be insured, otherwise you may encounter a classical “garbage in-garbage out” problem.
- A process model designed with the help of a process mining tool won’t necessarily be an optimal model. It is only the starting point for optimization.
- Mistakes or negligence residing at strategic or tactical levels are not identified within a process mining exercise. It will help to see a problem but not its reason.
- Process Mining itself offers no validation of a designed process model in terms of its alignment with business goals. Therefore there is no evaluation of how good the process model communicates business needs and how close it is to a selected strategy.
How to assess if Process Mining is the adequate tool for your processes?
Given all the above, we recommend starting Process Mining exercises only if the following criteria are met:
- A non-trivial number of process activities
- Model of the as-is process is not straight forward
- Event Log data is available and of sufficient quality
- There are processes in IT systems that do not correspond to a desired business logic
- IT architecture is cumbersome and complicated
- As-Is process model is already aligned with the company’s goals and serve all the business needs of the company.
About the Author
Pedro Ferreira Sales leads Be Germany’s Business Process Management service portfolio. He is a process management practitioner with many years of experience in both line management and project lead roles at systemically important banks – with a particular focus on generating measurable and verifiable added value through process optimisation.