In my role as Petri’s Community Magnate, I will be interviewing a variety of IT-savvy technology folks. Some of these will be Petri contributors; some will be tech-company employees; some will be IT pros. We will be tackling various subject areas in the form of 30-minute audio interviews. I will be asking the questions, the bulk of which we’re hoping will come from you, our Petri.com community of readers.
Readers can submit questions via Twitter, Instagram, Facebook and/or LinkedIn using the #AskMJF hashtag. Once the interviews are completed, we will post the audio and associated transcript in the forums for readers to digest at their leisure. (By the way, did you know MJFChats are now available in podcast form? Go here for MJF Chat on Spotify; here for Apple Podcasts on iTunes; and here for Google Play.)
Our latest MJFChat is all about business process mining. This chat is meant to introduce this technology and explain how it works. My special guest, the Chief Scientist of process-mining vendor Celonis, Wil van der Aalst, is a pioneer in this field. He discusses Celonis’ partnership with Microsoft in this space, among other topics.
If you know someone you’d like to see interviewed on the MJFChat show, including yourself, just Tweet to me or drop me a line. (Let me know why you think this person would be an awesome guest and what topics you’d like to see covered.) We’ll take things from there…
Transcript:
Mary Jo Foley (00:01):
Hi, you’re listening to Petri.com’s MJFChat Show. I am Mary Jo Foley, AKA your Petri.com community magnate. And I am here to interview tech industry experts about various topics that you, our readers and listeners want to know about. Today’s chat is all about business process mining. And my special guest is Wil Van Der Aalst, who is the Chief Scientist of Celonis, a process mining company headquartered in Munich and New York City. Wil is considered a pioneer of the process mining field with work starting back in the 1990s. And given Microsoft’s increasing focus on data culture, and the tools and services for enabling it, I think this will be a topic of great interest to our listeners. Hi, Wil, and thank you so much for doing this chat with me today.
Wil Van Der Aalst (00:54):
It’s a pleasure to be here.
Mary Jo Foley (00:56):
Great. So we’re gonna start out like a 101 level introduction to process mining. I know, I don’t know a lot about this and I’m guessing many of our readers may not as well. So let’s just start with the basics. What is process mining and what even is a process?
Wil Van Der Aalst (01:15):
I think since process mining is a very exciting technology, that I think can be applied everywhere. I’m more than happy to answer this question.
Wil Van Der Aalst (01:25):
So, I think everybody has a basic understanding of what the process is. So when I talk about a process, I’m talking about activities performed for particular cases. So an example of an activity would be a customer places in order. You make a payment in a hospital, somebody takes an x-ray. In a car factory, somebody assembles a car. And so these are activities. And I think everybody can imagine what activities are and these activities are typically performed for a particular case. And again, the word case sounds very abstract. But, if I give you examples, it’s clear. So a case could be a customer that is buying a product. It could be a patient that is being treated in a hospital, or it could be a car that is produced by a car manufacturer. So we have activities that are performed for cases. And the beautiful thing is that most of these activities in modern systems are being recorded.
Wil Van Der Aalst (02:29):
So if a customer places an order, that leaves a trace in the database of the shop where you’re buying something. If you’re in a hospital and you get an x-ray this is being recorded in the hospital information system, et cetera, et cetera. So these recordings are called events. So an event, refers to an activity, a case, and it happens at a particular point in time. And this is basically all the information that you need. So you can imagine that about most of the processes that take place in larger organizations, there is lots of data about events. And using process mining you can find out what is really going on, so you discover the real process. And this sounds very basic, but often discovering the real process is shocking for people that are involved in these processes. Once you have discovered the process, you can identify compliance problems. For example things are not working out the way that they should, you can see, let’s say where the bottlenecks, where there’s rework, and all these inefficiencies, and that is a starting point for improving it. So you can diagnose where a company could do better and you can automatically trigger, let’s say actions to do that. I realize this sounds all a bit abstract, but you can imagine that you can apply that in any organization of some scale.
Mary Jo Foley (04:03):
So if a company decides, okay, this sounds interesting. Why, would they institute process mining? I mean, couldn’t, they just figure this out without having a formal system? What can they get out of it, I guess I’m wondering.
Wil Van Der Aalst (04:18):
Well, what they can get out of it is, let’s say incredible savings. And also let’s say making their processes run much better. The problem is that if you look at an information system and you look at the tables, you have no idea whether the process is running well or is not running well. So many people have see just a small part of the process and they do not really see and understand the end to end process. So using process mining, you can uncover, for example, bottlenecks. So you may know that you’re not delivering your customers in time, but process mining helps you. What is the cost of that? If you have compliance problems, for example, you are delivering products without these products being paid, things like that. You can uncover such problems and the moment that you have uncovered them, you can also find out why they are, so why do you have these problems? What is causing that? So you see things like rework, what we call ping pong behavior, different people pass on work to each other without booking real progress. This is for example, people violating the four eyes principle, late deliveries, et cetera, eett cetera. These things are often very surprising and using process mining you can uncover that. And there are let’s say thousands of organizations all over the globe that are saving a lot of money by it, and also making their customers much happier.
Mary Jo Foley (05:58):
Are there certain sizes or types of companies that are more likely to benefit? Like, is this something mostly for big companies or can companies of any size and type benefit from process mining?
Wil Van Der Aalst (06:10):
So what you see is that the starting point is typically the bigger companies. So for example Celonis at this point in time has over 2000 customers. And these are typically the larger organizations like airlines, car manufacturers banks, insurance companies, et cetera. So process mining typically starts there. Because there, if you can save, let’s say 2%, often already you’re talking about millions of savings. But what is interesting is that process mining is becoming more and more the new normal. That you should do that. So in the larger organizations, you can see that first, they were focusing very much on procurement and sales processes, and they started applying it in all other fields. But also see, you can see that it becomes more and more accessible also for smaller for smaller companies. So in the end, I think it will for sure not be limited to larger organizations and it will become more and more like the new normal.
Mary Jo Foley (07:29):
Okay. Are there existing categories of products or processes that process mining competes with, or replaces, or is this just an adjunct to things that people already have in place around trying to figure out where they have bottlenecks and such?
Wil Van Der Aalst (07:47):
Yeah, so process mining is a completely new technology, right? It’s something that did not exist before. However for example, the success of Celonis has made many companies let’s say aware of the fact that there is a huge demand for these kind of capabilities. So what we see is that process mining is increasingly being added to existing offerings. So if you look at, for example the last year, you could see that IBM, bought myInvenio, SAP bought Signavio, Appian bought Lana Labs. So these were all, let’s say large deals, also a bit before like UiPath purchased ProcessGold. And this trend is showing that many of the let’s say, classical companies that were in either process modeling or automation, are let’s say seeing that they should add this to their software and are also adding it. At the same time you can clearly see the movement that for example, systems like Celonis are also extending the scope of process mining to also include, let’s say, low-code automation. So it’s something completely new. People that say, well, this is a kind of machine learning or something like that. That’s complete nonsense. It’s something very different from what companies had.
Mary Jo Foley (09:21):
Okay.
Wil Van Der Aalst (09:21):
But at the same time it will be the type of technology that will be added to, let’s say many, many existing systems.
Mary Jo Foley (09:30):
Okay. So I have some definitional questions here. Like how is process mining different from data mining and from predictive analytics? And then I also am curious how robotic process automation fits in, if it does, in here?
Wil Van Der Aalst (09:45):
Yeah. So if you look at the definition and that’s often the problem, if you talk about AI or you talk about machine learning then the definitions are always very broad, right?
Mary Jo Foley (09:58):
Right.
Wil Van Der Aalst (09:59):
People talk about machine learning and machine learning is, let’s say learning from data without being programmed. And then you could argue, okay, process mining is a kind of machine learning. And so from a definitional point of view, you could say, okay, it is machine learning, but now you look at what do companies that offer machine learning solutions. What do they actually do? And then they do something for example, related to deep learning, or neural networks, or image recognition, or natural language processing. And these are all things that have absolutely nothing to do with with process mining. If you look at the algorithms used in process mining, these algorithms are completely different from mainstream AI and machine learning. So, it’s very different. Also if you look at data mining, data mining, typically doesn’t look at process at all. So the notion of a process model, with a model that describes how processes are being executed is missing in all of the approaches that you just mentioned. So in that sense, definitional, it is a kind of machine learning. But if you look at the real technology behind it, it’s very different and unique.
Mary Jo Foley (11:20):
Okay. We have a listener on Twitter, whose name is Chris Hagan ask about the relationship between process mining and value stream mapping and management, which was something I didn’t know anything about, but I looked it up and it said, it’s a way to view a process flow from start to finish, and kind of analyze what’s going on. So are how are those two things related, if they are?
Wil Van Der Aalst (11:43):
Yeah, so there are many relationships because process mining touches the heart of any organization and things related to information systems. If you look at value stream mapping that comes more from the let’s say setting and the types of people that would normally also talk about things like Lean Six Sigma, et cetera. So you could think of process mining as a technology to support Lean Six Sigma, to support value stream mapping, but then actually based on facts. So there are these relationships, but again, it’s the same as with machine learning, people use these terms and then say everything is the same.
Mary Jo Foley (12:27):
Right.
Wil Van Der Aalst (12:27):
But I always challenge people, okay, show me now the capabilities to simply, I don’t know, extract data from Microsoft Dynamics, from SAP, from Salesforce, et cetera, automatically build the model, show where the deviations, where the bottlenecks are, and then you will see that these approaches that claim to support that actually cannot do that.
Mary Jo Foley (12:51):
Got it. Okay, so a lot of our listeners are very Microsoft focused. And so I’m curious, does Microsoft have anything in this space right now around process mining? I know they’ve got like Power Automate, Power BI, but like, I don’t think they have anything that would be considered process mining in the strictest sense, right?
Wil Van Der Aalst (13:13):
Well, I think we should distinguish between Power BI and Power Automate.
Mary Jo Foley (13:19):
Yes.
Wil Van Der Aalst (13:19):
So if you look at Power BI, it’s widely used, let’s say for data analytics. And Power BI itself does not provide any process mining. But what is interesting is that there is a handful of companies, I think the last time I counted, I counted six, or so. There are a handful of process mining companies that are building software on top of Power BI. And the most successful company until this point in time is PAFnow, who is doing process mining on top of Power BI. So this exists, at the same time if you look into the systems and you look at the actual engine, so the actual process mining work is typically not done in Power BI.
Wil Van Der Aalst (14:10):
So Power BI is the, let’s say is the interface that is shown to people, but the moment that you start running a process discovery algorithm, or the moment that you start doing conformance checking it is done outside of Power BI. So Power BI is like a UI to easily, let’s say do things with process mining. But of course, for companies that already use Power BI, this is very, very relevant and very interesting. If you look at Power Automate, it’s a bit different there, they recently added something that is called the Process Advisor, but the core offering of Power Automate is much more related to let’s say, low code automation, RPA catching let’s say triggers from the UI that the person is interacting with. If you look at the process mining capabilities that they released the first version last year, this is still incredibly primitive.
Wil Van Der Aalst (15:17):
If you compare it to, let’s say pure play process mining software. So there is a beginning. I think what is strong about Power Automate is again that it has the interface to, let’s say the user sitting behind the desktop, so it provides ways of catching signals. And if you want to automatically trigger something in a Microsoft application, you can use that very well. But it is doing very poor on what I would call the process intelligence part. So I think it is perfectly valid to combine a pure play process mining tool with Power Automate, with Power BI, but these process mining tools are much, much more powerful than the initial offerings in this space.
Mary Jo Foley (16:19):
Okay. I know a year ago, Celonis announced a deal with Microsoft where you were doing some integration around Power BI, Power Apps, and your product. I think Dynamics might have been part of that as well. So I’m curious where that’s going and how that’s been going for you?
Wil Van Der Aalst (16:36):
Yeah, so the CEO of Microsoft already also announced the collaboration with Celonis, and I know that the the Power Automate team is working closely together with the people in Celonis. But the way to view it is as I just described, I think from a process mining point of view let’s say a company like Celonis is lightt years ahead of what is currently let’s say, available in terms of process mining and Power Automate, that is really, let’s say very primitive. It is also that Power Automate is like the, as I said, you could call it the hands or the interface to the desktop itself. Not so much, so if I think about the typical scenarios that I mentioned earlier about, I don’t know, a car manufacturer finding out how to better produce a car. Like lot of organizations are using many different information systems. There is a lot of data let’s say in the back end. And let’s say systems like Power Automate, do not help you with that. It’s just, let’s say part of the story, but it’s completely valid to see that these things can, let’s say interact with each other in a very meaningful way, but then there would be many more examples for which you could say that.
Mary Jo Foley (18:02):
Okay, if you’re a customer who is not all in on Microsoft, say you have Salesforce, and you have SAP, and you have Microsoft. Which a lot of companies do, they have a combination of vendors they’re working with. Do you have to have a different process mining system for each of those platforms? Or how would that work from a customer perspective?
Wil Van Der Aalst (18:22):
No, that is definitely not the case.
Mary Jo Foley (18:24):
Okay.
Wil Van Der Aalst (18:24):
So you can think of process mining as a layer on top of these systems. So we, like I just saw a study of Forrester, and there it was stated that many processes touch at least 10 different information systems. So that shows you, if you really want to understand processes and you want to improve processes in terms of performance and compliance, you cannot focus on a single system. You need to have a layer that takes data from all of these systems, the different systems and integrates it. Because only then you can see the end to end process, and this is very important.
Mary Jo Foley (19:04):
Okay, got it. So for people who are interested in process mining, and they wanna keep up with what’s going on, I know they can go to the Celonis site and the blog, but where else would you suggest they go for resources and information?
Wil Van Der Aalst (19:19):
Yeah, so I’ve been working on the topic, let’s say now over 20, perhaps already 25 years. And in the beginning, there was nothing, but today there is lots of material that you can use. And so, there are several courses. So for example, a very lightweight course would be the course called “Process Mining From Theory to Execution”. That is a collaboration between Arden University and Celonis. And that’s a short, let’s say 10 hour course, where you see both the, let’s say practical things, but also more the theory behind it. Then if you really want to go deep there is the Coursera process mining course called, “Data Science in Action”. Interesting for me, is that it has been taken by over 150,000 people.
Mary Jo Foley (20:11):
Wow.
Wil Van Der Aalst (20:11):
So it’s very exciting to meet people all the time that took your course, let’s say years ago. So this is more for people that want to spend more time on it than the 10 hours. If you look at websites, one website that I think is neutral and provides an access point to lots of information about process mining, is very easy to recognize it’s called www.processmining.org. And from that you can find, let’s say pointers to books, courses, different software systems, different data sets. So if you’re new to the area, I would highly recommend to go to the site. And then in like 10 minutes, you have a very good overview of the entire area.
Mary Jo Foley (21:02):
Great. That’s a great resource, thanks. Well, thank you very much for your time and all the great information on this very interesting topic, appreciate it.
Wil Van Der Aalst (21:12):
Yeah, it was a pleasure to be here and to talk to you.
Mary Jo Foley (21:15):
Great. And for everyone else, who’s listening to this right now, or reading a transcript of this chat. I’ll be posting soon who my next guest is going to be. And once you see that, if you want, you can submit questions directly on Twitter using the #MJF Chat. In the meantime, if you know of anyone else or even yourself who might make a good guest for one of these chats, please do not hesitate to drop me a note. Thank you very much.