Why Simulation should NOT be in your Process Mining tool – What’s Your Baseline Shorts 3
The topic of this What’s Your Baseline Shorts podcast is Process Simulation and why you should not have this in your Process Mining tool, but rather as part of your design platform.
We speak about these topics:
- Why simulation?
- What do you need for simulation
- What are the outcomes of a simulation
- What are the different types of simulation?
A couple of related articles:
(The transcript is auto-generated and was slightly edited for clarity)
J-M: Hey folks welcome to this week’s What’s Your Baseline Shorts. I’m J-M Erlendson, a process and architecture expert, and I’m excited to be joined by Roland, who is going to tell us a little bit more about simulation today. Why don’t you introduce yourself first and then we’ll tackle today’s topic: Simulation?
Roland: Sure, hi I’m Roland. I’m the guy who had the idea for all those shenanigans, and obviously you can look at my bio if you want to see more details on that.
J-M: Wonderful, wonderful. Well, then let’s talk about today’s topic, because this is really What’s Your Baseline Shorts, the shortened version of our hour-long podcast. You can find on what’s your https://whatsyourbaseline.com. But today we’re talking about simulation; now it’s a question a lot of people have for us. How do I do something with the information for my process and Roland let’s tell more about what that means from a simulation perspective.
Roland: Oh well. I think simulation is more the idea that you can create futures (plural) in your tools and simulate, quote-unquote, what would be the outcome or to say it, in other words, you have a process.
And you change parameters on it and then you see what the different outcomes will be with the idea that you will then pick obviously the most advantageous and you use that for your implementation.
J-M: Yeah, well, there’s also another element of simulation that I really find important, which is future proofing. So yeah, so simulation is very good at creating those what-if scenarios so a bunch of like what if I change this thing. What would be my business value? So as you might imagine that’s used a lot for things like. Project justification says like, oh, if I just if I would only change these 3 things at the cost of a total cost of 2,000,000. I could save us a million dollars a year, like a huge amount of value, right? Um, but the other thing is that I’ve got a process today. That’s run, you know, 5000 times a week. What if tomorrow this became a core process for the company and it was run 25000 times a week?
Or 100,000 times a week, or a million times a week, where would it break first and so simulation can also be used to future proof processes. Um, so I feel like that’s a really important thing that people often overlook when they’re first thinking about the use case for simulation. But man, I have clients who are leveraging that today to do a lot of really good things.
Roland: And you open a can of worms with that, because simulation in itself is obviously not well. It is a thing, but it’s not a standalone thing. So typically it’s embedded in what I like to call the solution lifecycle, where you have some high-level analysis and then.
Roland: You might view simulation, say, as you described you know what happens if we go 20,000 times a day or you go to the next step where you say oh yeah, I’m designing a future state for one of our system implementation, for a reorg, you know, for a process improvement and you want to figure out what’s the most and advantageous there.
And then it goes through implementation and the execution, and then some vendors say, hey, we’re going to have that in our process mining tool, which is the measure phase, and we’re going to talk about that in a minute but um, J-M, tell me a little bit about what does a simulation do or what do you have to put in to get a meaningful simulation outcome.
J-M: That’s a really good question, and I mean, there’s kind of a minimum viable product, right? You need to have a demand model. How much are you asking? You need to have a process model, which is how it works, and you need to have some factor of time and flow to help you understand how long each step takes, and that could be formulas so you could have like a normal curve with you know the the parameters defined about each step, or you could simply have flat numbers like average times and then you can need to have routing numbers as in what percentage of time does it go down path A versus path B so that I can see how often it ends up in my different. Flow Scenarios and so demand time and routing that’s really the basic level.
Roland: So that’s that I think there’s a fourth one that you should put in, which is resources, typically human resources, right? How many people do this?
J-M: Yeah, I hope you do, but something you don’t have to, but it does a lot of good for you if you do. That’s correct. Yeah.
Roland: Simulate machines right? Think about some IoT thing or think about a manufacturing line. You know where obviously rechanging the jig in the machine for the next piece that you want to do is obviously also time consuming. You know, it’s the dynamic wait time and all these types of things.
J-M: Oh yeah, well if you want to get really down into it. You can even do things like simulating um, physical resources, so I can have like a hopper with a certain number of widgets in it and then it can track how often this process draws out on the hopper and how often replenishment comes to fill up my hopper again.
And so you can see those resources are what we use very heavily to find bottlenecks. Because oftentimes that’s where you’re going to find the most opportunity for improvement in terms of rebalancing loads and, you know, taking a look at future proofing. You know what demand will require in terms of resources on your side because, for instance, if you’re looking at a team of like 20 people and you get them in 5 different roles. We don’t have 4 people in each role. What you actually have is like 2 people in this role, 6 people in that role and 7 people in this role and 1 person in this role. If you switch around those resources, you could balance the work that’s being done much better, or if you switch around which resources are doing which task, you can much better adapt to who you have to be to make sure your process works better. So resourcing is a great way of understanding bottlenecks and then it gives you a really good way of helping to optimize around who you have.
Roland: So what is the outcome, then, of the simulation? Is it just a picture or what do you get? You know, you put all that effort in there.
J-M: Well, yeah, so ideally there’s a few different outcomes to simulation. The ideal is a simulation that is going to tell you or suggest to you possible paths for success based on your criteria.
Roland: How is that? Is a little person in there that whispers in your ear and says J-M you should do this or what do you have?
J-M: So number one is a lot of good simulation tools will write on your process steps different numbers about how they performed under the simulated conditions. You can use that to throw up dashboards to understand and highlight where bottlenecks are, where issues are coming up in your process, where things are needy or greedy process steps so you can say okay, so I’m gonna highlight the things in like in a bar chart. These 3 things are my biggest problems if I were to change 3 things in this process.
I suspect these would have the highest impact. The other thing simulation kickout for you is statistical files, things like Excel files, with a bunch of scenarios laid out and even scored scenarios. We take a look at things like optimization. You can score the combination of parameters that you can control to say this one has the highest net value to me. And so we we have a waiting system on each different parameter – like how much do I care about how long people are waiting versus how much do I care about how much money I’m spending on human resources. Like those are things you can weigh against each other and you can kick an excel file or some tools have an in-tool analysis for that.
J-M: Often have a visualization that you can watch, sort of simulate through and every step through to see. Okay, I have a problem here. It’s an outlier condition. What happened and let me watch back to people to understand how I have to adapt.
Roland: So that sounds super geeky to me which is one thing that we’re gonna talk about when we talk about the different ways of doing this But I want to push that a little bit out for a minute or so, but I think the takeaway is you as the potential simulation user should have a hypothesis. You know, you should know what you’re looking for. Are you interested in bottlenecks? How are your bottlenecks being defined? Is it just the step that takes too long, or is it the step that builds up a backlog.
And therefore it’s too long because you bind resources for this. So are you looking for the utilization – the age-old question. How many people do I need to do this job. You know,, so come up with an idea what your problem is that you’re looking at.
J-M: The other thing you can do even without necessarily having a hypothesis. One of the things we haven’t talked about a little bit yet is cost. A lot of organizations are trying to drive towards the minimum of cost. So it’s not actually just a hypothesis. It’s also like a goal in mind.
Roland: Right ? So that’s the first takeaway.
J-M: Like, what are you trying to get to exactly? So the only way you can get a savings model, and I actually do this with a really major technology company in the southwest. They said we have an external system that we’re using that is owned by a third party and every time we use it we use a system. It cost us money. How do we make sure that we use it only when necessary. So how do I have to reallocate resources. How can I re-engineer my process change around the steps that I’m actually doing so that we hit the system less with the same number of people and so cost was a huge factor so you have to include cost of steps. How much does a thing cost if it’s an external step or how much of our resources cost.
And then how long do those resources take so we can do a human resource cost. How much do our internal resources or our internal services cost? So if you’re looking at a chargeback or a showback scenario with one of your other business units or IT who’s providing the service to you.. We can do a pretty detailed analysis on where those costs are coming from. And then be able to minimize around that. That could be, instead of a hypothesis, that could be a kind of a goal. Yeah.
Roland: Makes sense, even though I would put it in the same bucket. You know, figure out what you want to know. That’s the second takeaway that you have, you can take that output right? So you got bar charts and pie charts and all that type of stuff you obviously could export it and I’m pretty sure if you have some lean person in your organization that will be your best buddies because then they can use this for their additional statistical analysis that is a little bit over the head for me in models like myself.
J-M: Yes, yeah, like I know, things like Matlab are really good for plotting statistical curves and like you can really get down to these first and foremost statisticians out there. We love you? No disrespect.
Roland: Yeah, if you go down the whole six Sigma rabbit hole. Yes, I’m with you.
J-M: But on this podcast we’re saying a little bit less complicated than what you’re doing, but thank you for doing what you do? Ah, yeah.
Roland: So the third takeaway is what I pushed out a minute ago as well. You need to understand what type of simulation you have and what I see is typically 3 types right? So the first type is what’s called discrete event simulation and that’s basically, I would say, brute forcing the simulation engine to do things. So if you say you have a process and it runs 3 times a day and it has so many resources and so many steps and each step takes so long and when you come to a branch, 80% go this way; 20% go this way this type of stuff. It literally creates 300 instances, and it runs those 300 instances obviously much faster than a human would do this, and it goes through and then creates those statistics while observing those parameters that we have this gives you the tightest control right? because you can trace down each individual instance you could say ‘Oh yeah, this is how this instance went and the other one went a different path’, right? Similar to what you see in process mining.
J-M: And this is really handy when you have the distribution of things on each step. So if I have a normal curve on that step, right?
Roland: Hold on, that’s different. That’s the second type, that’s the Monte Carlo simulation, where you go and you have probabilities and you throw individual samples on it, and if you do that often enough you obviously have the biggest enough population to come to a similar result. But it’s a method.
J-M: Well,Monte Carlo also relies on varying parameters like Monte Carlo you wouldn’t run the same simulation exactly as is with statistical things. You would also change the numbers in it each time you run it. So what discreet event simulation will give you.
Sort of how the process was performed in all of its different ways. It could have gone Monte Carlo in all of the different ways. It could have gone multiplied by all the different configurations of the process equals this huge chart of possibilities, and that’s when we do this the scoring as on Monte Carlo style simulation is when what combination of parameters led to the best outcomes.
Roland: Which means it’s faster because it’s just math, and computers are good at math, right? But you have less control, if you will. It’s not a brute force attempt. The third one is, and this is actually, the whole beef that you and I have about this topic, is when you see tool vendors. It’s particularly vendors in process mining. They will tell you they have simulation. They say oh we have simulation in our tool, but it’s more like a glorified what-if because, in process mining everything is based on the dataset that you put right from your runtime systems, which is basically your as-is information. So what you can do now is in these simulations is you can play with parameters. What happens if we had twice as many of these type of things, but it’s always based on the dataset. So what you’re missing when we think about the solution lifecycle that I just described before what happens if you want to throw out your whole process and do something completely different as part of your solution design. You can’t and that is I think that the biggest beef that I have with process mining where people say Oh yeah, we have our simulation in the process mining tool.
J-M: You’re screwed. Yeah, and process mining can inform the steps of your process for simulation. That’s a good thing having process mining as a data source but process mining isn’t the workplace. It’s the data source and when you start to take that and have that mental shift of like I can do everything in this one space.
Roland: Oh yeah, that’s true.
J-M: You can, but you are extremely limited on the choices you can make. We had a colleague we used to work with who used to say if I only have or if I take somebody a lot of work to do something, I only present one or two choices. But when I present 5 or 6 choices because I’ve got a lot more flexibility because of the tool I’ve chosen. Um, more likely to come to a better solution. So yes.
Roland: So let’s come to the end of this wonderful. What’s Your Baseline Shorts. What were the takeaways? Number one, J-M, come on number 1…
J-M: Number one is that simulation is something that you should be looking at from 3 different perspectives parameter variation, flow variation, and optimization of conditions.
Roland: Number 2 would be to have an idea of what you’re looking for right? Just don’t start out in the blue. And number three, J-M, is what?
J-M: Yeah, it is that your process mining tools are not your simulation tools. They’re a good feeder, but they are not a replacement, and so ladies and gentlemen. It’s been a pleasure as always.
Roland: That was so nicely said, and I’m Roland and we’re gonna see you in the next one.
J-M: I’m J-M: and we will see you in the next one.
Roland Woldt is a well-rounded executive with 25+ years of Business / Digital Transformation consulting and software development/system implementation experience, in addition to leadership positions within the German Armed Forces (11 years).
He has worked as a Team Lead, Engagement/Program Manager, and Enterprise/Solution Architect for many projects. Within these projects, he was responsible for the full project life cycle from shaping a solution and selling it, to setting up a methodological approach through design, implementation, and test, up to the rollout of solutions.
In addition to this Roland has managed consulting offerings during their lifecycle from the definition, delivery to update, and had revenue responsibility for them. This also included the stand-up and development of consulting teams, and their day-to-day management.