This paper presents a Discrete Event Simulation (DES) add‑in for JMP that enables users to model, analyze, and optimize complex process flows through an intuitive graphical interface. The add‑in provides a fully integrated GUI that allows users to construct simulations using configurable entity sources, queues, and processes. Entity sources support variable release timings, enabling realistic modeling of arrival patterns. Queues allow adjustable capacities, making it possible to represent bottlenecks, resource constraints, and system delays. Processes incorporate variable processing times, allowing users to capture real‑world variability and evaluate its impact on system performance.

At the core of the add‑in is a simulation engine that tracks each entity individually as it moves through the system, enabling detailed analysis of cycle times, waiting times, resource utilization, and throughput. Output includes a suite of graphical visualizations (such as time‑series plots, histograms, and animated flow diagrams) that illustrate system behavior dynamically, helping users intuitively understand system interactions and identify improvement opportunities.

The DES add‑in is broadly applicable across industries and operational challenges. Example applications include analyzing traffic patterns, evaluating paperwork or document flow, performing inventory and supply chain analysis, optimizing restaurant service operations, studying hospital emergency department flow, and many other scenarios where variability and resource constraints interact in complex ways. By combining the accessibility of JMP with the flexibility of discrete event simulation, this add‑in provides engineers, analysts, and decision makers with a powerful tool for experimentation, optimization, and data‑driven process improvement.

 

 

Hi, I'm Jerry Fish. I'm part of the JMP staff. I'm on a sales team. I've been with JMP for about 10 years. Today, I want to talk to you about a discrete event simulation add-in that I've built. I've written several add-ins over the years, and this is the most recent one that I've done.

What is a discrete event simulation? It's a model of a process that has things moving through the process, and we're going to call those things entities. They could be people moving through a checkout line at a grocery store or through turnstiles at an amusement park, something like that. They could be pieces of paper moving through a purchasing process at a business. They could be parts moving through a manufacturing environment. You see some of the other examples down there at the bottom, and we'll go through several of these as the talk progresses.

The add-in actually has three parts. First, we create the model, and I'll be showing an example of that. Then we run the model. There's not much to show there. It's simply calculations going on in the background. Finally, once the model has finished running, we can view the outputs, and we'll see some of that.

Within the model, we've got these entities that I described before. They're represented by circles. Each circle would be a purchase order or a person standing in line or a part in manufacturing. We've got sources that are represented by green plus signs. These introduce entities into the simulation, and they have the ability to handle different introduction rates. They could all be constant. Maybe a new person enters a fast food store, one every minute. That would be a constant rate, or you could specify a uniform distribution between a half minute and 2 minutes. It would give you a varying rate for the customers coming into the store, or you could have a normal distribution. All of those are available within the add-in.

Once those entities come into the simulation, they go into a queue. The queue is like a waiting room at the doctor's office or standing in line to check out at the store. Some queues, you might want to have unlimited capacity. Some queues, you might want to say, well, I can only allow 10 people in here at a time into this queue. We have that ability. A process then is where work is done on a particular entity, and that work may be something in the paperwork approval process for a purchase order. It might be taking information from a patient in a doctor's waiting room, something that has to act on the entity as they move through the process.

Then, finally, when they're out of the system, they could be exposed to multiple processes and multiple queues; they finally exit from the system. This slide shows a little about model creation. You're presented in the add-in with a canvas. This is based largely on the reliability platform and repairable systems' simulation, where you can drag these blocks onto the canvas. What we'll be doing, what I'll show in just a moment, is that we'll drag a source in, followed by a queue, followed by a process, another queue, and an exit. Once those blocks are dragged in, then we connect the blocks with arrows from one block to the next, so the software knows which direction the entities are moving.

Finally, those blocks, the sources, the queues, the processes, they can have parameters set up around them, for instance, specifying the rate that people come into the store from a source block. Let's start that animation. We're dragging in there a source block, a queue block, a process, another queue, and finally an exit. Then we'll connect those with arrows. Once the arrows are connected, we'll go to the right side of the screen, and we'll be able to edit those parameters. We can name the blocks. We can change, for instance, there is a constant distribution of incoming people, or here's a uniform distribution with a lower and an upper limit.

That's how we generally create a model, and then we would save the model, of course. Then we would run this model, and the model would produce an output file. Like I said, there's nothing really to see there. It just runs in the background. Finally, we get the output. I decided to display the output using the bubble plot function within JMP. What you'll see are these entities, these different colored dots that enter from the left and move through this very simple process. This is the simplest process that you can build with this current add-in. The source in this case has a constant release rate of one time unit per entity. They go into the queue. The process has a variable rate of addressing those entities, varying from a half a minute up to 2.5 minutes, I think. They then pass through another queue and finally out.

The queues, incidentally, handle the entities as first in, first out. The person that's been in the waiting room the longest is addressed first. Let's see what that simulation looks like. Of course, I've sped things up a little bit here. This is not real time. You also notice that the dots don't appear to line up right in the boxes. I've jittered those dots a little bit so they don't all land on top of each other, and we can see a little bit more of what's going on. You can see that sometimes there are a couple of people that back up in the Queue 2 waiting room or a couple of documents that back up that Process 3 is not able to handle, but generally, they're able to keep up with the incoming workload and not have things back up.

That's a very, very simple model. Let's take a look at another model. Example 2 is a little more realistic. This is a manufacturing line with rework. Here, raw materials or parts come in from the left. They go into a staging area, a queue, where they are acted upon with some work. The part is machined or coated or whatever that might be. They go into another queue, followed by an inspection. An inspector is here and has to decide whether the parts are good or not. In this case, I've said that 80% of the parts go on into this queue and then into the shipping department, but 20% of the parts they've got a blemish in the coating or whatever that might be. They're recycled back through here. They're prepped with some rework. Maybe they have to clean the coating off, and then they're cycled back in. Let's see how that looks.

Things only spend time in queues if there is something ahead of them in the process that's keeping them from moving into the next process. It looks like we're keeping up pretty well. There are a couple of parts in rework that haven't had the opportunity yet to get back into the standard workflow, but that's how a manufacturing process might look.

Let's look at a fast food counter. This is your normal fast food restaurant that you might come into. A customer comes in the door; they can line up in any of three queues. Each one has a cash register. We have one worker that's addressing the customers that come in, taking their orders. That order is then passed into this queue where it's taken by one of three cooks. The cooks prepare the dinner, the meal, the food, put it into this queue, where there is a bagger that takes that order and puts it in the bag and calls the customer. Then I don't have the exit box showing over here, but they pass that directly on through, and the customer can then pick up the food. That's the idea.

We've got variable customers, the spacing between customers that come in, the time interval. We've got various order sizes, so it takes different time to collect the order, and different times to prepare the order and then different times to bag the food. This is a realistic thing on how you might model a fast food restaurant. You can see here that our order taker is keeping up pretty well. Sometimes we have a couple of people that line up in the queue, but generally, the order taker is keeping up pretty well. Certainly, the cooks are keeping up fairly well, and the bagger is doing their job as well.

That looks good, but what happens, for example, when rush hour comes along, and we've got a much higher rate of customers coming in the door? I've left everything else, all the rates and things, the same over here. Let me see what happens for a rush hour example. Now we see that the order entry is taking some time. For the number of customers coming in, this single order entry person can't keep up, so you're getting some dissatisfied customers, probably. Maybe some people leave the store. We could model that as well. If the wait time is above too much, then we would have to recycle them out. That's the idea.

Still, the cooks look pretty good. What would you do to fix that? You might try putting a second order entry person into this system that maybe could relieve this, but perhaps that's going to overwork your cooks. These are the problems that you can address with this discrete event simulation add-in. There we go. The add-in is largely written. I've still got a few things I need to clean up. It will eventually be on the JMP Marketplace that you can download for free. I'm hoping to have it on there in time for our Discovery Summit, the in-person Discovery Summit in Baltimore from October 19th to the 22nd of this year.

The last thing I'll leave you with is the next time you're standing in a queue waiting to get on the subway or in a grocery store or at an amusement park or whatever, if that queue is long, and you're starting to get a little aggravated, start thinking in your mind about this discrete event simulation and whether the company has actually done things like this to model their system to try to reduce the amount of frustration for their customers. With that, I'll thank you for attending this talk.

0 Comments
Presented At Discovery Summit 2026

Presenter

Skill level

Beginner
  • Beginner
  • Intermediate
  • Advanced
Published on ‎07-15-2026 03:39 PM by Community Manager Community Manager | Updated on ‎10-30-2026 03:34 PM

This paper presents a Discrete Event Simulation (DES) add‑in for JMP that enables users to model, analyze, and optimize complex process flows through an intuitive graphical interface. The add‑in provides a fully integrated GUI that allows users to construct simulations using configurable entity sources, queues, and processes. Entity sources support variable release timings, enabling realistic modeling of arrival patterns. Queues allow adjustable capacities, making it possible to represent bottlenecks, resource constraints, and system delays. Processes incorporate variable processing times, allowing users to capture real‑world variability and evaluate its impact on system performance.

At the core of the add‑in is a simulation engine that tracks each entity individually as it moves through the system, enabling detailed analysis of cycle times, waiting times, resource utilization, and throughput. Output includes a suite of graphical visualizations (such as time‑series plots, histograms, and animated flow diagrams) that illustrate system behavior dynamically, helping users intuitively understand system interactions and identify improvement opportunities.

The DES add‑in is broadly applicable across industries and operational challenges. Example applications include analyzing traffic patterns, evaluating paperwork or document flow, performing inventory and supply chain analysis, optimizing restaurant service operations, studying hospital emergency department flow, and many other scenarios where variability and resource constraints interact in complex ways. By combining the accessibility of JMP with the flexibility of discrete event simulation, this add‑in provides engineers, analysts, and decision makers with a powerful tool for experimentation, optimization, and data‑driven process improvement.

 

 

Hi, I'm Jerry Fish. I'm part of the JMP staff. I'm on a sales team. I've been with JMP for about 10 years. Today, I want to talk to you about a discrete event simulation add-in that I've built. I've written several add-ins over the years, and this is the most recent one that I've done.

What is a discrete event simulation? It's a model of a process that has things moving through the process, and we're going to call those things entities. They could be people moving through a checkout line at a grocery store or through turnstiles at an amusement park, something like that. They could be pieces of paper moving through a purchasing process at a business. They could be parts moving through a manufacturing environment. You see some of the other examples down there at the bottom, and we'll go through several of these as the talk progresses.

The add-in actually has three parts. First, we create the model, and I'll be showing an example of that. Then we run the model. There's not much to show there. It's simply calculations going on in the background. Finally, once the model has finished running, we can view the outputs, and we'll see some of that.

Within the model, we've got these entities that I described before. They're represented by circles. Each circle would be a purchase order or a person standing in line or a part in manufacturing. We've got sources that are represented by green plus signs. These introduce entities into the simulation, and they have the ability to handle different introduction rates. They could all be constant. Maybe a new person enters a fast food store, one every minute. That would be a constant rate, or you could specify a uniform distribution between a half minute and 2 minutes. It would give you a varying rate for the customers coming into the store, or you could have a normal distribution. All of those are available within the add-in.

Once those entities come into the simulation, they go into a queue. The queue is like a waiting room at the doctor's office or standing in line to check out at the store. Some queues, you might want to have unlimited capacity. Some queues, you might want to say, well, I can only allow 10 people in here at a time into this queue. We have that ability. A process then is where work is done on a particular entity, and that work may be something in the paperwork approval process for a purchase order. It might be taking information from a patient in a doctor's waiting room, something that has to act on the entity as they move through the process.

Then, finally, when they're out of the system, they could be exposed to multiple processes and multiple queues; they finally exit from the system. This slide shows a little about model creation. You're presented in the add-in with a canvas. This is based largely on the reliability platform and repairable systems' simulation, where you can drag these blocks onto the canvas. What we'll be doing, what I'll show in just a moment, is that we'll drag a source in, followed by a queue, followed by a process, another queue, and an exit. Once those blocks are dragged in, then we connect the blocks with arrows from one block to the next, so the software knows which direction the entities are moving.

Finally, those blocks, the sources, the queues, the processes, they can have parameters set up around them, for instance, specifying the rate that people come into the store from a source block. Let's start that animation. We're dragging in there a source block, a queue block, a process, another queue, and finally an exit. Then we'll connect those with arrows. Once the arrows are connected, we'll go to the right side of the screen, and we'll be able to edit those parameters. We can name the blocks. We can change, for instance, there is a constant distribution of incoming people, or here's a uniform distribution with a lower and an upper limit.

That's how we generally create a model, and then we would save the model, of course. Then we would run this model, and the model would produce an output file. Like I said, there's nothing really to see there. It just runs in the background. Finally, we get the output. I decided to display the output using the bubble plot function within JMP. What you'll see are these entities, these different colored dots that enter from the left and move through this very simple process. This is the simplest process that you can build with this current add-in. The source in this case has a constant release rate of one time unit per entity. They go into the queue. The process has a variable rate of addressing those entities, varying from a half a minute up to 2.5 minutes, I think. They then pass through another queue and finally out.

The queues, incidentally, handle the entities as first in, first out. The person that's been in the waiting room the longest is addressed first. Let's see what that simulation looks like. Of course, I've sped things up a little bit here. This is not real time. You also notice that the dots don't appear to line up right in the boxes. I've jittered those dots a little bit so they don't all land on top of each other, and we can see a little bit more of what's going on. You can see that sometimes there are a couple of people that back up in the Queue 2 waiting room or a couple of documents that back up that Process 3 is not able to handle, but generally, they're able to keep up with the incoming workload and not have things back up.

That's a very, very simple model. Let's take a look at another model. Example 2 is a little more realistic. This is a manufacturing line with rework. Here, raw materials or parts come in from the left. They go into a staging area, a queue, where they are acted upon with some work. The part is machined or coated or whatever that might be. They go into another queue, followed by an inspection. An inspector is here and has to decide whether the parts are good or not. In this case, I've said that 80% of the parts go on into this queue and then into the shipping department, but 20% of the parts they've got a blemish in the coating or whatever that might be. They're recycled back through here. They're prepped with some rework. Maybe they have to clean the coating off, and then they're cycled back in. Let's see how that looks.

Things only spend time in queues if there is something ahead of them in the process that's keeping them from moving into the next process. It looks like we're keeping up pretty well. There are a couple of parts in rework that haven't had the opportunity yet to get back into the standard workflow, but that's how a manufacturing process might look.

Let's look at a fast food counter. This is your normal fast food restaurant that you might come into. A customer comes in the door; they can line up in any of three queues. Each one has a cash register. We have one worker that's addressing the customers that come in, taking their orders. That order is then passed into this queue where it's taken by one of three cooks. The cooks prepare the dinner, the meal, the food, put it into this queue, where there is a bagger that takes that order and puts it in the bag and calls the customer. Then I don't have the exit box showing over here, but they pass that directly on through, and the customer can then pick up the food. That's the idea.

We've got variable customers, the spacing between customers that come in, the time interval. We've got various order sizes, so it takes different time to collect the order, and different times to prepare the order and then different times to bag the food. This is a realistic thing on how you might model a fast food restaurant. You can see here that our order taker is keeping up pretty well. Sometimes we have a couple of people that line up in the queue, but generally, the order taker is keeping up pretty well. Certainly, the cooks are keeping up fairly well, and the bagger is doing their job as well.

That looks good, but what happens, for example, when rush hour comes along, and we've got a much higher rate of customers coming in the door? I've left everything else, all the rates and things, the same over here. Let me see what happens for a rush hour example. Now we see that the order entry is taking some time. For the number of customers coming in, this single order entry person can't keep up, so you're getting some dissatisfied customers, probably. Maybe some people leave the store. We could model that as well. If the wait time is above too much, then we would have to recycle them out. That's the idea.

Still, the cooks look pretty good. What would you do to fix that? You might try putting a second order entry person into this system that maybe could relieve this, but perhaps that's going to overwork your cooks. These are the problems that you can address with this discrete event simulation add-in. There we go. The add-in is largely written. I've still got a few things I need to clean up. It will eventually be on the JMP Marketplace that you can download for free. I'm hoping to have it on there in time for our Discovery Summit, the in-person Discovery Summit in Baltimore from October 19th to the 22nd of this year.

The last thing I'll leave you with is the next time you're standing in a queue waiting to get on the subway or in a grocery store or at an amusement park or whatever, if that queue is long, and you're starting to get a little aggravated, start thinking in your mind about this discrete event simulation and whether the company has actually done things like this to model their system to try to reduce the amount of frustration for their customers. With that, I'll thank you for attending this talk.



Start:
Tue, Oct 20, 2026 04:30 PM EDT
End:
Tue, Oct 20, 2026 05:15 PM EDT
Ped 1
0 Kudos