Transcription
You, hello everyone. Welcome to another QPR webinar. The weather today in your city is really beautiful. I guess spring is finally here. And not that everyone is feeling energized and ready for our discussion today, which is Process Mining for Process KPI Reporting. During the webinar, if you have any questions or comments, please leave them in the question box and we'll go through them after the webinar. Please also check out our upcoming webinars from our website. The next one will be released webinar for QPR versus Analyzer. And on the 5th of May, we will have Process Mining for Digital Transformation webinar. I can see that them was really ready. I'm excited.
Okay, okay, very, very good. So it's super great to see everybody again and and so so glad we have this big audience here. And I'm really happy to talk about Process KPI Reporting. And there's 29 minutes time left, so let me get started. Okay, so small introduction and then best practices around the three topics: how to get the KPIs, how to get the root causes, and how to make some predictions. So Process Mining is delivering much more than just the KPIs, but we will start from the sort of the basics and and go to the advanced topics in this presentation. And of course, questions and answers, we have reserved time. So just please write to the chat box.
Okay, so QPR, just a couple of words. As you see, more than 1 million licenses sold to more than 2,000 customers and 400 process mining projects, more than that. So it means a lot of process-related KPIs, and we have a lot of experience of different kinds of KPIs. And I'm glad to tell about our experiences there. So we are founded in '91, so soon we have been in this business for 30 years, and process planning has been 10 years.
Okay, let's start to talk about the KPIs. So here in the right-hand side, you see the KPI, Key Performance Indicator. So we are talking about something that companies use to evaluate the success of the organization. So you ask from somebody, "How is your business going?" Then you can say, "It's going good, it's going well, not so good." But if you want to be more concrete in your answer, you probably would take one KPI like, "Our sales is increasing," or "Our sales has gone down," or "Our profitability is good," or "Our customers are happy," or "The deliveries are being done on time." These are all examples of KPIs. And organizations have a lot of KPIs on a high level and also on the detailed level. And the key here in Process Mining is Process Mining is a data-driven approach. And with Process Mining, you get all the process-related KPIs very, very easily, as I will be showing in this presentation.
At the KPI reporting on another hand, that's the activity for defining what KPIs you want or had, then calculating the values based on the actual data that comes as a result of your business, then reporting those values to the relevant people in a relevant format, charts, dashboards, and then quite often also comparing against the target values. So the KPI very often has a target value that your company wants to reach.
Okay, and how do we do KPI-driven improvement? Well, as we discussed last week also about this PDCA Deming cycle. So in the planning stage, you analyze your processes, you define the KPIs, you see there are some problems, you find the root causes for those problems, and then plan for the improvement. This is activity where Process Mining is really, really heavily used. Then you do that improvement, whether it is RPA, Robotic Process Automation, or teaching people how to use the systems, or or or harmonizing processes, telling everybody that, "Hey, you should be doing like this." This step is like your business improvement. There you don't use Process Mining for that. You use Process Mining to plan your improvement. Okay, you use Process Mining here.
And the KPIs. Now, when you do your changes, then you analyze your KPIs, you see that what did change. So if we are failing in on-time deliveries, customers don't get our products in time, then you set the on-time delivery or on-time in full as your KPI. And here you are analyzing the KPI, "Did it get any better?" So you identify the changes in the process that were done here. How do they look in the data? Did it really, did these activities really make some change? And then the root causes for the improvements and root causes if if the KPI has gone worse. Okay, so this is you analyze, check against the KPI.
And then when you want to do this last part of the Deming cycle, you standardize your solution, you set the KPI targets. It can be quite an efficient way to set the target levels. They can even be tied into the salary and bonuses of the people. But anyways, you set the targets, you get kind of a traffic lights, green, yellow, red sort of signals where you are. And then you monitor continuously. That this is the KPI reporting kind of monitoring side. And then what we can do with Process Mining is that we can predict the ongoing cases. So instead of only looking backwards, as the KPIs are doing, you can actually predict on the potential future values of the KPIs. And more concretely, predict for individual cases what is going on.
Now, three main perspectives for Process Mining KPIs. Those three metrics are: the customer satisfaction, are your customers getting what they want? Are the internal efficiency perspective, are your processes working in a way as you want them to work, so that you are efficient internally? And then the third, the automate, the process automation, have you reached the desired level of automation in your processes? So these three dimensions are currently nowadays, we talk in 2020, these are the most important perspectives where organizations are taking Process KPI, Process KPIs: customer point of view, internal efficiency point of view, and the process automation point of view. And being good at all of these perspectives means that you have a very good process in place in your organization, like order-to-cash process or purchase-to-pay process. So you get a first-time right.
Alright, now let's get into the core of this presentation. Why should you use Process Mining for Process KPI Reporting instead of traditional Business Intelligence reporting tools? So obviously, Business Intelligence is an area where you create KPIs and reporting, and you report all kind of things. But there are three very, very concrete reasons why you should consider using Process Mining for the Process KPI Reporting. First reason: using Process Mining is much faster. You get the KPIs much faster with less effort. You really, really get the process KPIs fast. This is not true for all the possible KPIs there are in the world. You may have some financial KPIs that you use your financial systems to calculate those. But when it comes to the process KPIs, you definitely get them faster and with less effort.
Then we come into the Moody's. Automatically, you will get root cause analysis. This is something that the traditional Business Intelligence tools do not have. They don't give you the root cause analysis automatically. They they do not have that kind of functionality. I will be explaining why that is the case. But with Process Mining, you get it. And then you get case-level predictions. It requires a little bit of work here in the Process Mining area. But because we have the access to the root cause analysis, you get it like with just configuration. With Business Intelligence tools, if you want to make predictions on case level, you need to build a lot of stuff. You need data scientists to work with your KPIs and data warehouses and business data. But with Process Mining, you get all this for the process KPIs.
Let me go further to show you how. Number one: get the KPIs faster with less effort. So with traditional BI, you build the Key Performance Indicators manually. You take your data, and then you build the first KPI using your tools. When that is ready, you can move on, and you take the original data and you build your second KPI. Building a KPI is sort of a project in your organization. It may take anything between one day and 100 days, depending on a KPI. But there's always a kind of a project to build the KPI from the source data. With Process Mining, all the other youth first build the Process Mining model. That's the work you need to do. You need to build it because Process Mining needs it, requires the Process Mining model. That's the whole starting point. You can't do Process Mining without the Process Mining model. But once you do it, you already get the access to more than 1,000 standard KPIs, just like that. It's all there. You just point and click and see those. I will be showing how. So you get a huge amount of KPIs automatically just by building the Process Mining model.
So if you want to have an end-to-end process in good control, you can either start building KPIs one by one with the traditional BI, and when you read, let's say, 20 KPIs, you will have a lot of KPIs to maintain. Or you take the Process Mining approach where you build the model, and it will give you provide you all the standard KPIs as a package. Better yet, what is good is that, of course, you will have some custom KPIs that are depending on your, like, "I want to have a KPI for new product sales compared to the old products, new business among share of the total business." So then you need to define what is a new business, what is an old business. So you need to do some definitions. So with Process Mining and QPR Process Analyzer, you can define your custom KPIs, and you can with a much less time in the BI approach, with much, much less time, you can implement custom KPIs if they are not already included into the standard package. So you get the access to all the process KPIs, whether they are custom or not, with Process Mining. But you get them much faster and with much less effort. The KPIs with the Process Mining approach, you just need to build the Process Mining model first, and then you have a huge source of KPIs and data available.
Okay, here's an illustration that is showing it. So you may have the ERP systems over here, you may have the data warehouse over here, and you may create manually these KPIs, KPI reports. This is what is done in the traditional manual BI reporting teams all around the world. However, now for the process data, we have a new tool, much better tool, also for these teams who are building the reports, but also for the business people, process owners, process analysts, that can use themselves without the involvement of some IT IT personnel or kind of a to create the dashboard and KPIs. So with the Process Mining model, you get the standard KPIs just like that, they are already there, you just use them. And then you can create the custom KPIs from the process analysis model. And typically also these process models, they take the data from the data warehouses, but they also take data from low-level data from the ERP systems, or if you don't have a data warehouse, they can just take all the data from the ERP systems. Okay, so this is the benefit number one: you get faster and with less effort.
Benefit number two: you actually get the root cause analysis for all KPIs. How is this possible? Well, you could build in theory that kind of root cause analysis into the traditional BI. You could build a report with the KPI, and then you can build a root cause analysis that is having an access to all kind of extra data that it's correlating with that particular KPI that you had. But building this is a huge effort. Well, in Process Mining, when you build the Process Mining model, you already calculate more than 1,000 KPIs. So you are already creating a lot of more information that is needed only for this one KPI. So whenever you take one of the KPIs, look at one of the KPIs in the Process Mining model, you can have all the, let's say, 999 KPIs as your potential root causes. So you can see that with this particular KPI, what are the other KPIs that are correlating with this? Maybe they are causing this my KPI to fail. Of course, it requires you to understand that what is the cause and effect, what with what, what KPI is causing the other KPI to fail. But you have a system that has all the KPIs. So you have the root cause analysis ready over here, just like that. And if you build your custom KPI with a very, very specific rules, you already have the root cause analysis available for this, because it is using then always 1,000 other KPIs and features from the Process Mining model. So Process Mining model, the fact that you are building the model itself is a super good source for root causes. Anything that is included into the model is used as the potential root cause, our root cause details. And then once you go them one by one, you can keep on asking, "Okay, this is the most likely root cause problems in this region. Okay, what is it inside that region that is causing the problem? Okay, it is this product." And then you can go further.
So root cause analysis and how does it look like? For example, root causes for more than 10 weeks lead time. So case duration is more than 10 weeks. So you can just select these cases that are taking more than 10 weeks, or the KPI in this case is the lead time longer than 10 weeks. And then you see the root causes: region being Dallas, 50% of the cases are taking a long time. Customer group being kids, 21% are taking a long time, whereas the average was 15%. So this region Dallas and customer group kids are not part of the KPI lead time longer than 10 weeks, but they are some additional information that is in the Process Mining model that is in this scenario used as a root cause components. So you will get this automatically. You can present to your business users a report of a KPI and a root cause analysis. You get this. That's the promise. What you get from the Process Mining and the root cause analysis. You can, for example, see that, "Okay, you are getting 5% its product returns." That's now the KPI, product returns, 5% its level. And then you can see the root causes in the flowchart. You can see that within the customer pickup cases, for the percent of the cases that are being returned have gone through the customer pickup. And if it's not returned, then 16% have gone through the customer pickup. So it correlates that those who go through the customer pickup are more likely to end up into this being returned. So you get this root cause, a root cause analysis or the process steps automatically. That's provided with the tool. So this was benefit number two.
Then the benefit number three: since Process Mining has access to a lot of data regarding all the cases, the cases are going through certain activities and they have certain case attributes, that data can be used for machine learning-based prediction on a case level. And again, you could build the prediction capability with your traditional BI, but here you would definitely need a data scientist to start building that kind of stuff just from the BI point of view. You need more information than just to give you information to make a prediction. So you need to do something more than just this KPI number one. However, in Process Mining, again, as I say selling many times, the fact that you create the Process Mining model means that you have a lot of other information as well than just the one KPI. So you will have a lot of data in your Process Mining model. And now by configuring the case-level prediction feature that is already in QPR Process Analyzer, you can set up the case-level prediction to predict your KPI. So then you run that prediction for the open cases, based on the history of the old cases, completed cases, you make the prediction for the open cases. And this is available for you, and you don't need a data scientist for this. You need to configure it. You need to check what attributes and activities are correlating in a meaningful way with the result, so that you can increase the accuracy of the prediction. But you get the prediction. You will get the prediction also directly from the packets. But then you need a little bit play and work to improve the accuracy. But that's sort of a very much configuration work compared to building lists.
So that was the thing: three benefits. You get the KPIs faster with less effort. You get the root cause analysis, and you will get the case-level predictions.
So let's jump into the demo. I'm running here the QPR Process Analyzer user interface and let's see where I am. Okay, so settings. As you see, some of the KPIs are already visible here. So like the amount of cases going through this process, we see that 10,484 cases are going here. And 5% of the cases, 490, that I have the delivery changed. These are already KPIs. Somebody is using in the KPI reporting. The same KPIs, you can see in the flows, all the individual also one of the 99 cases called this route back to the picking done. Also, you see the durations, one day, 14 hours on average duration. And you can have these as an average and average with the weighted costs and so on. And then you can have the costs for those cases in the flowchart. The flowchart already provides a lot of KPIs. But let's move forward because this is this is sort of the very, very basics. But now let's let's let's take a bit of another advanced tool, which is the chart view. So chart view is the area where you build your KPIs. This is, for example, the case count, how many cases you have in total. And this KPI that we have over here, we can start working with this with this KPI. Let's see if we, for example, want to benchmark the regions. I could say that, "Okay, now we see how many cases we have in different parts of our organization." Let's say that I have in Chicago 1,559 cases. So now let's use the root cause analysis right away. Let's say how these cases could be different than the other cases. So we just click this, set it as the root cause criteria. And now we see immediately from the flowchart that the cases in Chicago who are more likely to go to the customer pickup and being returned by the by the customer. And also in this root cause analysis for these case attributes, we see that of course, the cases from Chicago, as the top line, 100% its are from Chicago. But the second most important root cause, account manager is Mary Wilson, and the happy customer attribute is no, so they are not that happy. Whereas on the other hand, some other people, reasons, they are happy, and the region is New York. So whatever you design as KPI with this, with the chart view, you can do right away these root cause analysis.
Let's look at some other kind of KPI. Now, let's change another dimension. Instead of doing this kind of comparing regions against each other, let's look at the case duration. So now we have the case duration in weeks. So again, the whole data, and this is the amount of cases per weeks. So now if I look at these long-lasting cases, what do I see here? When I put this to be the root cause selection, I see this root cause is showing the region Dallas, 55% of the cases are taking this long time, 10 to 34 weeks. And in region New York, they are much faster. So I have the root cause analysis here. And then I can also see the root cause analysis on top of the process flowchart with the colors. So the long-lasting cases tend to go more often to this "Purchase Order Created" in, whereas the quick cases are going to this "Shipment Sent" route more often. So whatever I design as a KPI in this environment, where I have the measures and the dimensions that can be used immediately for the root cause analysis.
Well, we can go forward. I was just using the very basic case attribute and the case duration. But of course, I can take a bit more advanced cases, where an event occurs before another event. Let's think about the on-time delivery. I could say that if, for example, my this confirm delivery date is first, and only after that we make a shipment, then that could be considered as a problem. So if the confirm delivery date is first, I have said that I will be delivering next week Monday, and the shipment is taking place after that, then I failed my on-time delivery promise. So those that fail are now in this "Yes" category. Well, I have now configured my KPI using these graphical dropdowns. I have the "Yes" and "No" over here. So now I just go, let's take away those that limitation. Well, I have 179 cases where I failed my KPI. So I just click on this and show me the root causes. So now I see the root causes that with the red color, that those cases that that that tend to fail the KPI in the "Shipment Sent" route, there are much more of those cases, whereas in the "Invoice Created" and "Payment Received" routes, there are less of those cases. And I can take all my data from the Process Mining model and and see that, "Okay, if it goes to picking, it's again more likely to to be failing this KPI." So this is root cause analysis on top of the closure, on top of the flowchart. And then the root cause analysis using the case attribute of the model. With the same data, it gives that this was a KPI I designed, and here we see that if the account manager is William Davis, and 13% of the cases are failing my KPI, and region LA, 8% of the cases are failing the El the KPI, whereas some other areas, 0% are failing. I don't, on average, only 2% are failing this KPI, which we just together in this webinar built using the the the chart view settings that have already all these KPIs here. And now comes the point that if I want to turn this into your very, very custom KPI, I just I just convert this for custom editing, and now I see it here, and I can go into this custom editor. So if I want to start creating very, very custom specific stuff, I can just go here and use the person editor to edit this expression, and I'm getting to be whatever whatever I want. But I still, I will still be able to use the root cause analysis also for this custom custom expressives. So let's let's get this this back. Okay, so so what we have here is an environment for making the KPIs. And then if I have some KPIs already here, let's take this case attribute and let's say select one of the, for example, the KPI "Automation Delivery Completed" KPI. So then if I have already created pre-created KPI values, then just clicking the KPI value obviously gives me the root cause analysis. So I can just I can just run the root cause analysis for automation delivery complete, it's easy, no completely delivered. Then those cases tend to be more of the product group genes, supplier genes, international region Chicago, whereas the other KPI values come from here. And you also find the KPIs for the happy customer. So when you have set up your system, and then when it seems like, "Okay, you are having challenges with the customer customer happiness," then you have the KPI which is showing that happy customers, you have 7,300 cases, and here unhappy customers, 3,000 cases. Again, if you want to see the root causes for the unhappiness, you just keep this one and pull it as a root cause, root cause criteria. And immediately here you see that the region Chicago, 63% of the customers are not happy, whereas the average not happiness is 30%. So Chicago, Mary Wilson, account manager, Robert Miller, and if your internal foe was not successful, that is correlated with the end customer being not being not happy. So this is a setup for for creating the KPIs. Much more functionality. I would love to book a meeting with you. I hope to do it myself or somebody else from QPR, and we could show you how we can create any process KPI that you have in your business, no matter what KPI it is. We can show you how to, what kind of Process Mining model is needed, what is a process ID, and that what kind of data is needed. We can build a small sample model to show you how it works, to show the root causes also for your KPI.
Alright, the the the last two minutes I want to show you the case-level prediction. So case-level prediction, as I said, is is a functionality that comes together with the product. It's over here, and it's used in a way that you have a particular KPI. Here we are predicting something to take more than 30 days. And then let's say that I want to make that prediction for those cases that had not yet been been completed. So I haven't got the payment yet. And then let's see, let's give a little bit of source data for this prediction. And now what I see here is that 511, 511 items in my process are predicted to fail this, to take more than 30 days. So I'm giving, I'm having a KPI here, I'm selecting which attributes I want to give to the prediction, and what are the events that I want to give to the prediction. If I, for example, take away all the other events, we can see that, let's say like here, so we get a new prediction that our 307 items or still to us. So with this new prediction, new data, 307 items are predicted to take more than 30 days. So I can customize these selections here. And I have the full access to the machine learning tools to check the accuracy of my predictions. And once the accuracy becomes to be good enough, then this list of the cases that you have over here, here like these are individual cases that where you haven't received the payment yet. So these lists can be given to the people for some allowing these pylons that they can see that, "Okay, what other cases that they predict that the money will not come in time?" Okay, but it can be any SLA, service level agreement, or on-time delivery. But what other cases that will be failing the on-time delivery? So this is a case-level prediction feature.
Okay, that was the demo for today. I hope you remember this: getting the KPIs faster and with less effort. And in addition to that, with Process Mining, you get the access to the root cause analysis for whatever KPI you build, and you get the case-level predictions. There is a fully functional dashboarding system. So when you then build your KPIs, you get everything in as excellent business graphics as the reporting tools, because this is the same kind of toolset that the reporting tools are using. But the major difference is that underlying these KPIs, you have the whole Process Mining model, which is capable of giving you the root causes and the predictions. And it gives a lot of those process KPIs for free. So you just build a model and you get them automatically.
Okay, here's a list of the webinars. So we have done for today. And let's see, let's see what kind of questions there are. Okay, how is it always so difficult to open? Okay, can you choose the case ID? So case identifier is one of the most important things parameters in Process Mining. And definitely, when you are working with QPR Process Analyzer, let's say for example, in this like in this case-level prediction view, we were getting a list of the cases. Let's see if we get a list of three splice 30-days prediction. And let's take some of the data so that we can actually make some predictions. If you don't give any date, okay, so here is a here is a case ID. So you immediately see that when I click on these, sorry, when I click on these individual cases, like this one, the flowchart over here is now showing this particular case. So in this particular case, it has gone so far that it has sales order created and picking has been done. It hasn't got the payment yet, but it is predicted to take more than 30 days. And you can also see that in the cases analysis, you can see all the all the details for for any particular case, like in this particular case, right? I just select it over here. So you see you see the details, and you can even see the older case attributes for live images, case attributes. If I show all the case attributes, then I see the all the properties for this case. So so over here. So yes, definitely you can always drill down into individual cases level.
Okay, is the list sorted? Well, yes, if you are concerned about sorting the lists, I would say that in this prediction prediction list that we have over here, so you can always sort these these this information. Let's we could maybe replace this with something else. It's quite a lot of information filling up here. You can just sort these by clicking the headings. And of course, when it comes to when it comes to making your chart views and sorting the elements in your KPI reports and dashboards, that is studies as well, of course, possible.
Okay, and then there is a question regarding the KPI. So how do you make a KPI for conformance? That is a very good question. So let me so let me show you how to do it in QPR Process Analyzer. So here we come into this area of how do you make some particular KPI. And it's good to each good to remember that that we could have we could be showing you a lot of possibilities how to do this. Let me take so we we would be happy to show you how to create any any particular KPI you have. But let's let's first look at this one. So here what we have is a conformance analysis. So we have a process flowchart that is coming from the data, and then we have a design model. So this is what how the process should go. This is how it should go, and then we have the reality. Now, one definition of a KPI is, "Please give me a KPI that is telling as a percentage how much of my business is going according to this process." Well, you can give that specification to a Business Intelligence consultant working with the reporting tools. You can give this BPMN flowchart and say that, "Please give me that KPI." But it's more like 100 days to build this BPMN logic of how the case should go forward. But of course, with Process Mining, you don't need to do that because it's a built-in functionality. So now we see immediately here the conformance trend, 47% of the cases are conforming with the design model, and it's the green area here. And you see the violations, the variations, and you can even see the root causes, as you can see for any KPI in the system, you can see the root causes also for this. Like in cigar, 84% are non-conforming. So this is a good example of a Process Mining related KPI that comes directly from the packets. And on the other hand, is very, very, very time-consuming to create in the process, sorry, in the Business Intelligence systems. And these conformance KPIs can be embedded into your into your your dashboards.
Well, here we have count of non-conformance cases by this supplier, for example. So there are certain suppliers, and then you see how many cases they had. All over the green supplier has this many non-conforming cases, and these are the sort of the the non-conformance reasons. And here are the cases, kind of other charts helping you to get into the details of the conformance phenomenon in your business.
Okay, let me see if there's. Yes, there is other questions as well. Okay, okay. Hello, can your application take information from multiple platforms, SAP, CRM, and and other files? Yes, definitely. That is the point. That's absolutely the point in analyzing end-to-end processes, so that you get the data from very many different sources. And let me show you just very, very quickly this page in the in the QPR. So QPR website. So this is the QPR connectors page, which is showing you ready-made connectors to order to to our systems, other systems like SAP or all Salesforce and so on, Microsoft Dynamics. And then there's a blog article, which I am quite happy with, because I be in the team to write it, "Real-time Process Mining." So the point is that there's a fully functional system to connect to your data systems, extract the data, and transform it into the Process Mining format, and then then load it into the memory, so that it can be analyzed. So so this transforming the data, this is the heart of the Process Mining. This is where the Process Mining model is being built, and we do that very systematically, and we get all the KPIs and root cause analysis for free.
Okay, to analyze information. So yeah, without having, yes, you don't need to switch to platform. You you load, you can load all the data you have in all your systems, manual files, and everywhere. You can load all the data, and you should load all the that you should load all the data from all the different sources to the same end-to-end process. So obviously, you may have some may product plant maintenance data in some system, and you may have some financial data in some other system that you use in some financial processes, and then you may have some quality-related data in some quality-related system, ticketing systems, and auditing all the auditor processes. So you may have many systems, and they are not all providing data to the same processes, but you will have many end-to-end processes like order to cash, where you get data from from very, very many systems.
We were talking about the case-level prediction. Oh, I see there are so many questions. Okay, yes. All right. I think we've used the time. I'm super happy you were participating again also this week. This was the third in our series of the Process Mining use cases, and there will be three more use cases to go. So we have now covered the RPA, robot process automation, process improvement, and now this Process KPI Reporting. Next week will be our release webinar, and then after that, it's going to be the digital transformation, IT development, and the auditing and compliance. And we are very, very happy to have a meeting with you to discuss one-to-one about your particular case, maybe build a demo model together for your data, or discuss about the Process Mining roadmap. Just please keep in touch and book the time, and let's see how we can we can do things together. Thank you. Thanks. Bye.