On Process Mining. Q&A with Anastasia Migunova, Senior Data Scientist, Deloitte

Q1. What is process mining? 

Process mining is a technology that provides detailed insights in company’s processes. Nowadays information systems store a lot of data about events happening in daily business. The goal of process mining is to create a process model based on this data which will be used in business process analysis. It enables compliance checks, supports process redesign and decision making. Areas of application are extremely wide including financial processes, supply chain, production and others.

A distinction should be made between process mining in research and business fields. Process mining in research field concentrates mostly on algorithms for establishing a process model. However, only a small part of these algorithms is used in business projects, which are focusing on applications of the created process model.

Process mining can be divided into 4 steps: data extraction, data modeling, process visualization, process analysis.

Data extraction. Data for future modeling and analysis can come out of different sources: SAP, all kinds of data bases, Excel-files and so on. On this step, it is identified what process attributes are available and relevant for future analysis and can be linked to the process model, as well as filter criteria which should be applied to the data set.

Data modeling. During this step connections between different systems and documents are being established and data is being homogenized (e.g. if the process is partially covered in SAP and partially in any other system). It is crucial to identify “process anchors” – documents which will serve as a source point for linking further events and attributes. Usually only one document type is chosen as an anchor (e.g. if one decides payments to be anchors in purchase-to-pay process, purchase orders without a payment will not be shown in the model). In Deloitte we implement our own Process Mining Framework which allows to choose as many anchors as needed.

Process visualization. The established process model is uploaded in one of process mining tools (Celonis, ProcessGold, LANA Labs, PAFnow, Minit and so on). The existing tools usually provide users with possibilities to see process flow and calculate process relevant KPIs based on the model.

Process analysis. Once process visualization and implementation of desired KPIs are finished, business users overtake developed dashboards for continuous process improvement. Later, further adjustments to the process model and dashboards can be made.

Q2. What are the main challenges in process mining? 

Find a compromise between business requirements and technical limitations.
 In big corporations with many subsidiaries huge amounts of information are being recorded. It is not always possible to represent an end-to-end process with all its variations (e.g. there is a limit for a number of unique events – around 600). A common decision should be found with business users to meet technical requirements and still to be able to provide an extensive process model.

Homogenize processes conducted in different systems. Although many companies use SAP, different parts of a process could be covered in other systems with a different data structure. It also happens that foreign subsidiaries are using their own tracking systems. In such cases it could be challenging to bring all the data to the same structure which is required for process model creation.

Define business hypotheses and goals for a process model. As said before, there are certain limitations to what can be shown in the process model. Therefore, it is important to create a clear plan on what is expected to be present in the model and what process anchors should be used.

Q3. What are the main lessons you have learned in using data analysis? 

First of all, it is important to understand that data analysis itself will not solve your problems. It is just a tool which could help you to identify the root cause. You will still have to find and implement the solution to your problem.

Second, you need to have a plan for data analysis. There are enormous opportunities and often a lot of data waiting to be analyzed. If you do not have any idea what you are looking for, you will most probably find nothing.

Third, data quality is crucial. It is better, to save time and money and not to do any analysis at all than to do something with bad data and make a wrong decision based on that. If the data is collected and is planned to be used in future, it should also be maintained as well as possible.

Fourth, you should understand the data you are working with. It helps a lot in order to avoid mistakes and wrong conclusions.

Q4. What kind of mathematical modelling do you use? 

For pure process mining implementation it is not essential to perform any mathematical modeling. SQL scripts are used to aggregate data and create a valid process model.

However, there is a vast field of process model extensions using mathematics and analytical techniques. For example, probabilistic models are used for predicting delivery times, credit risks, make-to-order cycle, etc. AI is being implemented to predict next process steps, which should simplify future intelligent process automation.

Q5. You have a PhD in Mathematics. How did it help in your professional career ? 

It helped me to reach certain level of analytical abilities, to learn to overcome obstacles and to find my passion in my professional life. So, it is not a title which matters, but the entire way you have to go for this.

Mathematics is a universal tool and areas of its applications are continuously growing. It is used in science, industry, finance, sports, medicine, politics, linguistics. So you are never bored and can always find something that excites you.

Qx  Anything else you wish to add? 

There are a lot of techniques and tools for leveraging process optimization and business improvement. Process mining is only one of them. But every tool requires work and devotion to reach the final target. There is no perfect solution for everything, just find the one that suits your needs better.

Anastasia Migunova is a Senior Data Scientist in Deloitte Center of Process Bionics who helps organizations with implementation of process mining technology. Before joining Deloitte, she was a data analyst in Procter&Gamble, where she used mathematical analytical tools in industrial environment to support decision-making process. Anastasia completed a doctoral degree in Fraunhofer Institute for Industrial Mathematics while applying her scientific results and modeling know-how as a basis for companies’ research projects.

You may also like...