I presented a series of webinars around Process Mining together with Anne Rozinat from Fluxicon about Process Mining.
The first webinar was held at October 3rd and some very interesting questions from the attendees were raised. This blog post aims to provide answers to those that could not attend or that continuously had doubts about the possibilities of Process Mining.
Question: Imagine that a company has a BPM system and have many process running on that platform. Also this company has a BAM System to analyze the process. how can we introduce Process Mining integrated to this technologies ?
Answer: I’m going to expand the question from BPM system to any other system that model, design and implements business processes even those that are brought to life on real time (instance mode). There is an assumption that the process model that was implemented reflects 100% the reality as it occurred. That is absolutely wrong. There is no straight line from start to finish. It’s not a matter of exceptions occur, it’s a matter that work does not happen the way it was designed for.
Even if the process is highly structured one instance can stop and halt at some point, other can stop in a different activity. There are no equal instances. This means if you look to process execution as a heat map, you will find patterns, practices that differ from the “main success” scenario.
If the process provides degrees of freedom, or is designed and executed on instance mode you can compare if the social network is aligned with the process type. In this situation knowledge type is exploratory or exploitative, but if the social network configuration among the people that participate is strong central with a network star, knowledge flow is inappropriate and a solution will take much more time to achieve, because the flows are being centralized by a manager rather than the people that need information to act.
BAM (Business Activity Monitoring) tell you everything about how many requests are being handled, how much time does it take, workforce balancing, if the company is missing the the service level agreement, but does not shows you WHERE you are lagging, WHAT is the cause of missing the targets. process Mining is process oriented information. Thus, it shows the pain points and bring you awareness HOW to act in real time.
Question: You presented some methods to model the control-flow perspective (using the order of activities), but what can you say about the data-flow perspective? There are also techniques to model the data, to see how data changes during a process?
Answer: What Process Mining technology does it’s to reconstruct for each instance the execution path (the process model) and sum it up to deliver it with all the expectations, loopbacks, and unexpected paths that the process took in reality. Each instance has predefined data attached that is defined at the start of the project. Data can be transformed in analysis dimensions (roles, costs, and data). Thus if the focus is to understand how data flow occurs, it can be analyzed under that perspective.
Question: (I love this one because is out of typical process automation domain). Suppose there are 1000 steps in the assembly, and 3 are in the wrong order. Would I be able to find this using your tools without manually checking every detail?
Answer: Imagine you are analyzing an assembly line that fills up bottles of wine. The assembly line has built in sensors that records the operations, the errors, and calculates the OEE (Overall equipment effectiveness). Once Process Mining is agnostic, you can analyze the recorded data and figure it out in the millions of wine bottles, those that went into label placement twice. Thus it’s possible to discover duplication, because there is an error setting up the labeling machine and you need to reprocess the label task.
Question: Can we use process mining to simulate or analyze interactions among business processes. i.e. the points or activities in the processes that can cause negative effects in the business processes (multiple organizations or same organization)?
Answer: Sure. Today (finally) processes are not being recorded only only inside corporate boundaries. They exist and return back to the company that sparked it (initiating an order that goes to a supplier warehouse, then a forwarding agent, gets back to the company, the supplier sends an invoice and hopefully there is a payment). Once there is a common key across multiple systems, it’s possible to mine the processes and get answers independently the company that performed some activities.
Question: The processes you demonstrated today are lasagna type processes. There are on the other hand also processes that have a lot of variation in process flow and little numbers of instances. e.g. in Healthcare. What kind of help can PM give in these types of processes?
Answer: Contrary of some opinions that darkness can see the ray of light if the process is structured, there are lots of REAL world applications under Adaptive Case Managment. Here are some pointers:
- Application of Process Mining in Healthcare – A Case Study in a Dutch Hospital – R.S. Mans, M.H. Schonenberg, M. Song, W.M.P. van der Aalst, and P.J.M. Bakker;
- Beyond X-Raying a Care-Flow: Adopting Dierent Focuses on Care-Flow Mining Filip Caron, Jan Vanthienen, Jochen De Weerdt, and Bart Baesens;
- Finding Structure in Unstructured Processes: The Case for Process Mining – W.M.P. van der Aalst and C.W. Gunther
- Monitoring Unmanaged Business Processes – Nirmal K. Mukhi
Question: In what phase of a project would you recommend process mining? Is it important to have a certain level of CMM with steady and well knows processes?
Answer: Maturity models can be a useful to help people to make decisions, BUT, we can have a scenario that most of our processes reached level 5 (maximum of the scale) and need to be modified because management sensors, or customers vote that the process will not satisfy the requirements and expectations. Thus once again Process Mining can help to find here the process must be adapted.
Question:Can process mining be used in a continuous manner? Is it still interesting beyond initial use?
Answer: It can be used after change occurs and resume if changes were effective and the predicted outcomes were achieved. Is just like taking two CT scan, comparing if the cure eliminated the decease.
Question: I’d like to use process mining in order to find error causes in a manufacturing environment. The cause of these errors is often located in the details. How can you prevent abstraction away from details too much when creating a process model?
Answer: With smart filtering it’s possible to look at the needle in the straw. I mean you can look at a particular category in a short period of time, look only to two activities in hundreds that occur in the process, look into two roles only …
Question: How is your experience to get the data you need to analyze processes? Do companies provide the data easily?
Answer: As data become an asset, its analysis and efficient exploration is important for companies to compete effectively. However the fragmentation and the volume of available data make it impossible to analyze manually. Today most of the enterprise systems have tools to extract data from databases, like knowledge views. I remember for example that BMPS databases are impossible to understand at extract data from it. Using tools like knowledge views are handy and easy to do it. SAP is known to have difficult to understand database structure and we have experience extracting millions or records using ad-hoc query. One data attribute that is needed to process mining is the activity name. Activity names are not stored in systems with beautiful names like, created, progress, completed. Typically is stored state transition codes like 0, 10, 20, etc. If your IT dept. cannot provide help because there is no system documentation, initiate a new process instance and monitor how the state codes are recorded in the database and reverse engineering it.
One final thought: don’t let data greed to start playing a role in the project. Don’t extract data you don’t need to.
Question: Process Mining can analyze 100% of the cases, in an auditing context. How can you be sure that 30% of the cases are missing? Exactly the 30% that were fraudulent?
Answer: Some techniques can me use to cross check records. For example compare using a BI system the cases count with traces in the database, specially the ones that ere deleted.
Question: Can we apply the process mining techniques on a data of any Transaction Processing System? For example: on data generated by ATM?
Answer: Process mining can be used in transaction systems like the ones used by front office, when a customer pays a bill, request energy connection or pays goods at the supermarket. It can show for example information about how the customers like to pay for transactions (cash, check, credit card) that combines with the process execution sequence that help managers to make decisions about customer preferences, or find causes that halts process execution (I just remember, transactions can take much more time at a given period of the day due communication infrastructure, and thus must be improved in order to not increase customer waiting time).
Question: What is the difference between the data mining and process mining.
Answer: Data mining is data-centric, not process oriented. I’m not saying data mining is useless. But does not show where and why organizations are for example performance lagging. Business intelligence combined with data mining shows how many complaints were made, how many were answered and what response time, by category and other attributes (correlates important things like complains started by women take more time to handle, but if an offer is made to compensate the problem it’s more likely that the churn rate decreases) but hides where people waste time or perform more tasks beyond that necessary, repetitions, duplication, etc. On the contrary, the process mining shows managers where they must act to achieve the performance expected and can act in real time, because they know in the part of the process that must be, for example accelerated or resolved.
Question: We have already implemented a BPM system. How does Process Mining fit to our BPM system?
Answer: BPMS record every interaction. Users log in and log out, every piece of data of all process instances, error logs and in some configurations used in military / defense or pharmaceutical industries every change that was made to data stored by the system (for example a user changed a test value from 4,0 to 4,3). Those interactions are recorded using what is called an audit trail (see image bellow). In this example you can see information about a process instance. By default it has all the data necessary to use process mining. Thus process mining tools integrate easily with BPMS.