100 Years of the Control Chart: JMP Live Brings SPC to the Masses (2025-US-30MP-2553)

Hello. My name is Kevin Gaffney, and I have the honor of being able to present to you my case study in using JMP Live and the value that my organization has found JMP Live bring to us. The story begins with statistical process control deployment using JMP Live, and then the additional value that we saw. How are we using the control chart 100 years after Shewhart's original presentation?

I'd like to thank my co-authors, Ari Gomez and Brian Moritz from Medtronic. They were the manufacturing engineering team on this project. And Wendy Tseng, the developer, and Alex Lippincott, our JMP admin, who was instrumental in our pilot project.

I've been a JMP user for almost 25 years. I started my career at Motorola, and I was introduced to JMP 2, I believe. Throughout my career, I've held various roles as Development Engineer for process, for design, and now I support operations. I've seen both sides of the development of process and how we use statistics in the development phase, and then how that is transferred into operations.

I'm passionate about helping engineers on the operations side really understand how to utilize the great development work that is transferred to them and how they can actually control more of the value that they bring to the organization by using tools like JMP Live, as we'll describe in our example.

What you're going to hear in this talk is first the control chart demonstration in JMP Live. As the name of our presentation suggests, we started our journey exploring a better way to involve SPC in our day-to-day activities on our value stream. We found that the new offering in JMP Live really provided more value than just a simple control chart. We'll explain how that's done.

By the time we wrap up, I hope to show you that there's additional value here beyond just the engineering needs of monitoring your process, which is accelerating engineering outcomes, avoiding software proliferation, and minimizing infrastructure burden across the value stream. All of these value propositions are things that I believe a manufacturing engineering team can hold the key to, and not need a large support team to really get the value out of their data.

As we said, this control chart has been around for 100 years, and you can see here that there is an example of the first control chart from Shewhart, and this was presented by William Woodall in a JMP white paper just last year. You can see there in the graphic that the key is the time series data.

In a slightly later version of the same information, Shewhart shows that you can understand your process better by going from aggregated, simple histogram view of data, into understanding your subgroup behavior in a time-ordered basis. When you put on statistical limits for trends that are statistically unlikely, then you can actually start to take proactive measures and maintain your process with higher value.

Now, in today's age, what's happened in 100 years since this original formulation is that there's been more statistical rules added to common SPC platforms. JMP has 14 additional statistical-based rules that you could turn on and check for in addition to the original Shewhart control, which is based on plus or minus 3 Sigma from the average. But in this day and age, we also expect that the control chart is accompanied by an out-of-control action plan that provides rigorous actions for the entire manufacturing team to take if they were to see one of these violations in the statistical control of their process.

Now, building on this was a formulation that Donald Wheeler presented, that he called Process Health. Process Health is an understanding of your manufacturing line when you combine not just the SPC control chart, but also the process capability. When you understand the state of your process with respect to, is it in control, yes or no, is it capable, yes or no, you have four distinct risk categories to your process.

The ideal state means that you're delivering product to your customer within their specifications according to your CPK or PPK metric, and that you have nothing but common cause noise in your line, and there's no special causes or statistically unlikely trends in your data. That's what we would call the monitoring state.

If you say no to both of those, meaning that you can't deliver what your customer needs, and you have many special causes in your data, you're in a state of chaos. Both of those states are fairly obvious, and we don't like to be in the state of chaos, but when we're there, we know it.

The other diagonal shows two more tenuous states to be in. The first is when you're not delivering a capable process, and you're not meeting your customers' requirements all the time, but you may be stable. In that case, at least a business can plan predictably and what they need to do to continue to deliver at least the quantity of material that your customer needs.

But for me, as a manufacturing engineer, the scarier state to be in is a brink of chaos. This is where I may have an acceptable PPK or CPK value, and I think everything's going well, but I'm not really in control. At any moment, I can have a special cause jump up and shut my line down. That's a really hard state to be in.

What we're going to talk about as we show the JMP Live story is that part of the extension of the value view of JMP Live beyond just SPC, which is only one axis of this formulation, is that we can simultaneously also look at the other axis, which is process capability, without ever having to leave our analysis tool or even our data set to go find the answer of both of these viewpoints.

Where does the additional value come in? If you work in a large manufacturing organization like I do, then the entire engineering workflow is usually supported by cross-functional teams. Cross-functional teams also usually come with silos, and those silos usually come with inefficiencies and costs.

This isn't a perfect formulation for every engineer, but most engineers could probably break down their workflow into about 10 steps. I'm not going to go through all of them here. We're going to focus mostly on the right-hand side, which would be more towards the manufacturing end. We're going to overlay how we use the combination of JMP or JMP Pro and JMP Live to show extra value in how a manufacturing engineer could actually own many parts of this workflow without having to get lost in cross-functional program management. They have the power themselves to execute the majority of their work and drive additional value to the company.

At this point, I'm going to transition over to the actual JMP Live portal, and you'll be able to see our pilot study. What we see here is the portal for JMP Live. JMP Live is an on-premises database, so your company would own the database and the administration of it. JMP Live is a software application that sits on top of that. Again, that's great news for me as a manufacturing engineer, because that puts the power of administrating the tool closer to my value stream than maybe an on-cloud tool that is owned externally. This is very familiar. You see here at the top, it's just a web address that the engineer would get, and then they're able to see whatever analysis that they have access to within the JMP Live ecosystem at their company.

The other thing I want to point out here is I am not using JMP Desktop to see this. As a manufacturing engineer, I may not even be a JMP user. Certainly, if I was a technician working on the line or an operator, I definitely wouldn't need to have JMP Live. I'm sorry, I wouldn't need to have JMP Desktop. And if I wanted to make this part of a management review or a broader global team, those team members also don't need to be JMP users. This is truly published through a web portal that anybody in the organization can have access to.

Now I'm going to tell you about my perspective on how I use this as a manufacturing engineer. When I'm talking from the perspective of Wendy, who is the development engineer on this project, or Alex is the administrator, I'll point that out. I would start my day, ideally as a manufacturing engineer, coming to this web portal. I might not even have to leave home yet to get a first look at how my value stream is running. I could go to the website, I could check out my value streams of interest.

To do that, while there may be a lot of analysis available to my organization, my particular value stream here would be presented to me here. We call it the discovery value stream, but it's contained here in a drill-down folder structure. I would come here and the first thing I would see is before I even drill down into the analysis that I have what looks to be a warning within my analysis. That's interesting already, and really driving value in some urgency in what I want to do the first thing in my day.

I'll drill down into my reports that I work with my development engineer, in this case, Wendy, to produce. I see that her and I have collaborated so that there are four reports that were interest to me on a daily basis. You can see now I have a warning indicator over here, but now, because I only have a few analyses presented to me, I can pretty quickly see that the warning here is on this one control chart. If I had many, many analyses within one window, maybe I would start here to see all of those.

This is a very simple example. I'm going to drill down on this graph first because not only is it my control chart, but this is also showing me a signal that I need to understand quickly. I drill down to my SPC chart, and I see a few things. The first thing I see is that I have a legend here where there's some additional information that I might see on my control chart. In addition to the raw data, I would have indicators if there is an open signal, a signal that's being investigated, or a signal that's been closed. I don't see it here on this chart, and so I might click over to warnings and say, "I want to quickly drill down to that warning without needing to hunt for where I should be spending the first part of my day."

I click down into that warning, and I see here that panel opens up that gives me the detailed information about the warning. You see here on the right-hand side, I have a control chart on a process input called current, but when I look at this panel, I see that the first indications of a warning are actually on an input called time.

What my analyst has done for me is very cleverly and pleasantly taken my six process inputs of interest and actually put them into a tabbed format. Now I don't have to leave a nice portal viewing to see all the information that I know I'm going to be interested in at a moment's notice. When I click over to my time input control chart, now I can see that I actually have several open signals, and they are related to violating the Shewhart control, which is the upper control limit.

That's interesting and important for me as a manufacturing engineer, but what I'm going to show you next is really important for driving efficiency into the organization. I told you that 100 years ago, maybe SPC was driving value by simply having time-ordered data and a plus or minus 3 Sigma. I also told you that now my organization expects us to have a rigorous out-of-control action plan. JMP Live allows me to get a head start on that expectation without leaving my data analysis or my reporting tool and going to another system.

I can go right into this panel here, and I can click on the actual individual signal. Now you can see that I'm highlighting this first violation of my control chart for whatever time period or shift that this was updated on. I can go right in, and I can see this was row 116 of my database. The actual value was 3.21. This is what the violation was for. We talked about there's actually 15 violations available in JMP. There could have been other reasons for a violation here. Now I can go right here, and I can indicate whether I am investigating or whether I have closed an observation. I can assign the action to someone. I can actually leave a note here, and I can look at the history of what this particular signal has seen.

This is really important. When you remember that this is a web-based tool that anybody can see that doesn't need to have JMP Desktop, now I can have an entire organization participating in my value stream statistical process control with one single tool. We've brought SPC to the masses because now they don't have to be a JMP expert or a statistical expert or a manufacturing engineer. We're driving value because A, I, as the accountable engineer, am drilling down quickly within minutes to get the organization to a more stable state. Now the rest of the support team can participate with me in the same portal without needing to go to multiple different tools. We're reducing software proliferation and reducing infrastructure.

I'd mentioned also that beyond SPC or process stability, we want to understand what is the true Process Health of our line. I'm going to go back into this Process Control window, and I'm going to show you that using the multiple different platforms and tools that JMP has and the way that you can mix and match them to create whatever visualization you want. I know that the very next question I'm going to ask by management is, "You have some instability in your process, and we saw that here with time, but what is your capability?"

I can embed in the same report a quick summary of process capability so that I can get a head start on answering those questions, and I can understand if there's a big risk to the conformance or scrap rate of my process. This is all process input. I'd go from an indication that there's an instability in my process input to understanding, what is the capability of that input? But then what I really want to know is what is my product level impact to these variations.

Typically, the way we get at that is through capability charts. You might have a situation where you have some histograms of your actual product output, and you know what the historical distribution is or the historical capability. You might go here and look at, like I said, these typical process capabilities, summaries, and get an idea if you've gone from, say, what was once a normal distribution to something that's now bimodal and maybe edging towards one tail or the other.

In this case, now I'm still bouncing between multiple different tools. I've had to go to multiple reports. Even though I'm still in JMP Live, which is great, I've done a lot of clicks to get to understand my Process Health. Is there a better way to do that? Now we start to see the layering of additional value here that JMP Live brings beyond SPC.

What I worked with my data analyst to do is say, "I'd really like to see all this stuff at once. I really don't want to have to bounce between multiple different reports." What Wendy has delivered to me is a combination of all the analytics that I might want. Now I see them at my fingertips. JMP actually provides Process Health as a platform in their software. I don't even need to go create special code or write script to get this idea of Process Health and these four states, it's built into JMP. What my analyst has given me now is a way for me to see process capability and process stability, like we talked about in the four quadrants.

Now, what we're looking at is my process outputs, which are thickness and roughness, which is ultimately what the business cares about, is making sure we're meeting our customers' expectations. I can see the Process Health from my customers' point of view here, and I see that I'm in an area of being stable but uncapable. I can see the time-based frequency of my SPC signals, which we saw before in the actual SPC chart. Now I can start to look at my history of capability. I see my stability as well as the differentiation between my long-term and short-term capability. Now I'm starting to build the efficiency into my workday.

I still wanted to start with those control charts because, as a manufacturing engineer, that's my organization's preferred way to manage our line. I quickly want to be able to drill down and look at the bigger impact of my manufacturing variation on my customer. I can do that here.

Again, keeping with the tab theme, I can also look at the same perspective for my inputs, and I can see I have a lot more signals recently on my process inputs. Now the process inputs are on the left-hand side of this axis, and I'm looking at the number of signals over time, and I can see what is the Process Health of all of my input channels. All of this is really adding value to my day as a manufacturing engineer.

When I saw all of this from an SPC point of view, I was happy. This was great, and I became an advocate for JMP Live. The real value that I think most engineers would get excited about is within JMP Live, I maintain interactivity with most of the tools that are out there. If you're a regular JMP user, especially if you're a process engineer and a JMP user, you've probably grown to love the prediction profiler.

Hopefully, you're actually using the simulator and some of the defect simulations, et cetera, that JMP has. What I've done here, I've worked with my analyst now to go take some of that very rich development history that was available in my company's technical reports, and we mined those for the critical transfer functions that were pinning together those two outputs I showed you of thickness and roughness and those six inputs.

The company's knowledge base has these transfer functions. Now I can bring those transfer functions to JMP Live in an interactive way. If you're not familiar with the profiler here, what we have is a multi-response optimization situation where I'm looking at my outputs, and I can interact in a continuous way and slide any setting for these inputs to a region of interest, and I can see how it impacts not only the one output, but both input simultaneously.

It also has some statistical uncertainty in there. In this case, we have a lot of data, so the uncertainty of our estimates are low. There is an uncertainty band here that you can see around any one of my point estimates, so that I'm able to also get an idea of the uncertainty of any optimization that I'm interested in.

If there are particular settings that I know our business would prefer to run because of some type of business constraints, then I can just go into the profiler a different way and I can type in those exact settings that I would prefer to run. I can then look over here on the Y-axis and see what is the point estimate of my resulting outputs as well as the uncertainty tolerances.

This is all great for me as the manufacturing engineer. If you're a development engineer or a data scientist, and you want some more information on the models, all of the information that you would typically get with the modeling steps in JMP Desktop are also available to be shared. Again, remember that I'm viewing all of this information outside of JMP Desktop. This is all through JMP Live. I don't even have to have JMP Desktop open. I can now start to share my analysis, if I was a development engineer with my team from a different perspective. Adding value to the different types of engineers, even though my story is from a manufacturing engineering viewpoint.

That, to me, as a development engineer or a manufacturing engineer, is an unbelievable value to me. Within just a few minutes, I've gone from coming in and seeing what is the current state of my line through a web portal that I can get through anywhere, all the way through optimizing scenarios without ever having to leave the web portal or open up a different tool.

Now I'm going to tell you a little bit about the nuts and bolts of how JMP Live works. We talked a little bit about the warnings over here. If you're a JMP user, you know that there is contextual menus depending on where you're at in your analyses. If I was within a particular analysis, and here our SPC chart, and I was to go over to the data, I actually have access to the data table that built that analysis, and so it's not just a visualization with nothing behind it. I actually can drill down without ever leaving the web portal.

I can also start to see a little bit more information about what is the history of this chart. When was it updated? Who was updating it? Who has access to it? I can leave comments so that we can have a chat history here if there's interesting things that I want to share with my team. Then this is the panel that shows how do we actually create these charts, and so this would have been an automated control chart that gets updated.

The way that's done is that you can manually upload data to the JMP Live server. You can have it done automatically via a script. If you were to do that, then you would paste your JMP script here. You could then have reference tables. You could assign credentials if there was a firewall or if your refresh script required certain credentials. If you use JMP Add-Ins, and JMP Add-Ins were part of your analysis, you could attach them here. Then the refresh schedules is where you would actually, just like in a calendar type setting, say how often you want to update your data.

If you are a JMP user, and you would like to get access to the data and maybe extend the analysis, you can come up over here on the right-hand side, and you can see that there's open in JMP. This would open up the data table and these reports that you're seeing in a JMP project. Then you would have to be a desktop JMP or JMP Pro user, but it would seamlessly open up all that at your fingertips.

Then there's typical site builder or web-based tools, there's bookmarks, there's sharing capabilities, and then there's the ability to manage what we call the posts within the screen. You can see here, here's an example of where we can see the regeneration strip. That's what we call the automated updating of the scripting, it's here. We're not going to talk about scripting or workflow building here, but if you're not a scripting expert, don't worry. Even my role as a manufacturing engineer, this is easy for me to do.

If you use Workflow Builder and JMP, it creates the JMP script along with your just regular use of JMP through the drop-down menus and creating the visualizations you want, even going into the Dashboard Builder. All of that is captured automatically, and it's just a cut and paste into this window here. If you're not a JMP scripting expert, you still have all the power at your fingertips through your normal use of the JMP tool.

Publishing is easy. Again, I'm not going to show it, but literally, it is a drop-down button within the JMP Desktop, which is simply to publish to JMP Live, and so with a couple of clicks of a button, you are able to publish all the reports you want to your workspace, which is the JMP Live word for a portal with focus reports.

That brings us to the end of our JMP Live demonstration. I'm now going to go back to the slides and wrap up. I hope in the very brief example that you've been able to understand what is it that we as engineers are trying to accomplish, all the way collecting data through optimizing our line in about 10 different engineering steps. We can accelerate those engineering outcomes through JMP Live because now we have the ability to share our analysis seamlessly with a larger organization without every user or consumer of the analysis having to be a JMP user or JMP expert.

I've shown that in my organization, we've been able to avoid software proliferation through choosing an SPC platform that not only does SPC, but also allows me to use that same data set and do a richer set of data analysis all the way through statistical optimization. That same JMP Live tool allows me to do some of the administrative work I need to do in terms my out-of-control action plan without ever having leaving JMP Live, which is unbelievable.

I'll stop here. Just a reminder of where we've been and some of the engineering steps that we need to take, and that the combination of JMP and JMP Live really gives an extreme amount of power to the engineers at the end of this workflow who are accountable to optimizing their engineering outcomes to the company. Thank you.