COMPANY CONFIDENTIAL. DO NOT DISTRIBUTE.
Data and statistical science. Applied to 3M.
3M’s vision is to apply science to life; JMP provides invaluable statistical capabilities to deliver the innovation that makes that vision a reality.
_____________________________________________________________________________________
Challenge: Navigate an increasingly siloed, competitive marketplace by deepening product knowledge and responding more efficiently and directly to customer requirements. Transform business processes – both manufacturing and transactional – by making them more repeatable, efficient and customer-focused. Better leverage the advantages of 3M’s diversified portfolio by adopting a more data-centric culture.
Solution: Early adopters of JMP at 3M see the software as an efficient, cost-effective platform for running state-of-the-art analytical applications with quantifiable impact. Among those practices where JMP is considered an essential tool are research and experimentation, customer studies, materials testing and process improvement.
Results: With JMP, engineers at 3M have gained the ability to:
- Construct detailed, data-driven profiles of customer needs.
- Streamline experimentation processes to more efficiently deliver on targeted customer requirements.
- Optimize manufacturing operations with built-in quality solutions and enhanced reproducibility.
- Grow the organization’s culture of Lean Six Sigma.
_____________________________________________________________________________________
3M creates a multitude of materials: abrasives, adhesives, ceramics, composites and films, to name just a few. Its capabilities span an array of processes and functions, from modeling and molding to polymers, radiation, replication and more. And its offerings serve a diversity of industries: biotech, imaging, filtration, medicine and optics.
3M’s ability to deliver on its innovation promise across such a diversified portfolio involves an extraordinary amount of product, process – and most importantly, customer – knowledge. That’s why engineers across the organization rely on JMP®, a robust analytics software package specially designed to help engineers navigate even the most sophisticated R&D and manufacturing challenges facing the industry today.
Data-driven pedigree
“As we evaluate new ideas and develop new products, we depend on tools like JMP to support experiments and understand the results,” says Michael Sykora, a product development specialist in 3M’s Commercial Solutions Division. 
“When you're in research and development, you want to play around with the data to answer the question, ‘What am I actually learning here?’ I might want to plot y versus x, and then switch out the y, and then bring in a second y,” adds his colleague Carol Kramer. “And I want simplicity, like drag-and-drop to see what a chart will look like.” The fact that JMP offers both powerful analytics and a dynamic, user-friendly interface is one of the reasons JMP is so invaluable, she says.
Kramer approaches data visualization and analysis from the perspective of someone well-versed in statistics. An advanced quality engineering specialist in 3M’s Separation and Purification Sciences Division,
she brings to the table a background in industrial engineering, a keen interest in the mathematical modeling of processes and certification as a Lean Six Sigma master black belt.
“My interest is in using data to make good decisions,” she says. “My fundamental philosophy is about how we build manufacturing and transactional processes that are efficient, repeatable and that meet the needs of the customer.” As she explains it, “all work is a series of processes, and all processes have variation. We’re trying to reduce that variation to achieve a more predictable outcome for customers.”
Like Kramer, 3M has a long history of statistical thinking. In part that’s because many of the industries 3M serves, like electronics and biopharmaceuticals, demand statistical rigor. But this philosophy can also be attributed to former 3M CEO Jim McNerney, who originated Six Sigma within the company – a commitment continued by current CEO Mike Roman, himself previously a Six Sigma director. And in part it’s driven by 3M’s focus on ISO standards, from ISO 13485 for medical devices to ISO 16949 for automotive to ISO 9001 for quality management. As Kramer points out, the ISO 9001 standard specifically notes that “methods to analyze data can include statistical techniques.”
For every problem, its own solution
In working with data, both Sykora and Kramer leverage multiple tools. Sometimes that’s JMP, though not always. But, they say, JMP sets itself apart with unique features that can’t be matched by other software.
“At 3M, we want to give people the tools they need to make the most efficient and effective decisions, Kramer notes. “And every person thinks differently. So we don’t want to replace every other tool with JMP. But we want to make available the tools to help them make the right decisions and document why that decision was made.”
Sykora agrees: “I like Excel as a platform for large data analysis. But if I'm moving results in and out of statistical analysis tools, I find JMP is much easier to use.” He especially likes the ability to add experiments in a subplot approach. “When you do an experiment, JMP allows you to add more and more data,” he says. “Other software doesn't allow you to do that in a statistically significant way.”
Using JMP means less time and effort spent planning the experiment upfront and anticipating the results. “With JMP, each experiment is less complex than if I had to design a full factorial experiment to capture all the response surface I want,” Sykora explains. “It’s a lot less time and cost to build a screening experiment on a smaller design space and then reiterate within that design space or around an adjacent design space.”
Kramer, for her part, believes software like Minitab can be appropriate in some situations. “Minitab is written for people who think like statisticians,” she says, “whereas JMP is more for engineers and visual learners.
“But comparing JMP to Minitab is comparing apples and oranges,” she’s quick to add. “JMP is much more powerful. As soon as I saw the data visualization in JMP and capabilities like Graph Builder, Chart Builder and Fit Y by X, I was sold. When I saw you can have categorical data, and you can have variable data, and you can pull the menu over and play with the data – it was truly a take-my-breath-away moment where I said, ‘This is so powerful, I need this.’”
Much more than plain vanilla
Sykora offers a good example of a capability that simply wasn’t available before JMP. His group develops films such as window treatments, glass finishes and the coatings used in electronics and automotive applications. Sykora’s division often presents customers with a range of newly developed films to get feedback on desirable product attributes. But asking for input on a large set of samples can lead to imprecise responses.
“Before we had JMP, we’d present customers with a stack of films and expect them to know what we were asking,” Sykora relates. “It was like asking, ‘Compared to vanilla ice cream, what does this film taste like?’ They might respond, ‘This film tastes better.’ Then we’d ask, ‘What about coldness versus sweetness plus five other attributes?’” The qualitative responses were difficult to measure.
“We’d rank it on a scale of 1 to 10, but the resolution on a 1-to-10 scale is meaningless,” he points out. “Should it be 1 to 10? Should it be 1 to 5? What if an attribute is so horrible that it's worse than 1 or so great that it's better than 10?”
The solution was maximum difference scaling, or MaxDiff, an effective way of understanding perceived value. “With MaxDiff in JMP, you don't need to worry about whether your scale is correct,” Sykora says. “The MaxDiff ranking tells you what you need to know. And because it’s a nonlinear ranking, you can assign a clear winner.”
Just as important, 3M can combine those results with analytical measurements. “Let’s say customers preferred a film with a certain color,” Sykora says. “We can plot the MaxDiff ratio against our analytical results and use that to set the specification.”
Achieving data purity
There are numerous other ways that 3M leverages JMP. For one, 3M offers a series of internal trainings in design of experiments (DOE). The last training in the series is an advanced course that uses JMP, because JMP is the only software that provides advanced DOE capabilities like definitive screening designs.
Kramer’s group, which services the biologics industry, uses JMP in several ways: test validation, manufacturing process improvement and R&D:
Test validation
Kramer’s group runs reports to validate its test methodologies. In one case, a report initially failed gage repeatability and reproducibility (R&R), a measure of the amount of variation in a measurement system.
“We had to troubleshoot and figure out why,” Kramer recalls. “So I looked at it in JMP, which has the ability to do honest gage R&R. I found that if we had used JMP from the start, we would have passed the criteria the first time. In fact, we would have saved six weeks repeating the test-method validation.”
Process improvement
3M had made a significant process improvement in the manufacture of a filter for purifying active pharmaceutical ingredients. Kramer needed to test samples from the new manufacturing process. But testing involves cost in both time and materials. Kramer wanted to limit the sample size while yielding statistically justifiable results. She achieved that goal using Exploratory Data Analysis (EDA) in JMP.
“We were sampling every 100 feet,” Kramer recalls. “So I modeled the sampling in JMP by going to 250 feet and then to 500 feet. All I did was simply drop out the data. JMP showed that there was no chance we would go out of specification because of the process improvement. It gave me visual data that justified reducing the sampling plan.”
The benefits were tangible. “Our test capacity was maxed out at that time, so reducing the sampling plan actually helped us avoid a back-order situation,” Kramer explains. “That’s a significant outcome in terms of meeting customer needs.”
R&D
3M innovates a lot of advanced materials. Prototypes can be costly to produce. That places a premium on efficient, well-designed experiments.
“JMP is crucial to advanced DOE at 3M,” Kramer insists. “For example, some of our chemists needed to design a test for a new filter medium. They came to me and said, ‘We have five levels of pH and four levels of another variable, and we want to see the design space and where the curvature falls off.’”
Kramer describes the simplicity of the solution: “I immediately opened the Custom Design platform in JMP,” she says. “I wasn’t concerned that it was categorical data. I put in the variable levels and the number of runs. In five minutes I hand the chemists sheets that showed how to run their tests.”
Benefits are just the beginning
The use cases of JMP at 3M demonstrate specific advantages. But the statistical discovery tool also delivers broader, longer-term benefits that are just as important. One benefit is the potential for improved decision-making. “At 3M, our decision-making needs to be data-driven,” Kramer says. “So we need an effective way to analyze and visualize data. And once you start using the visual capabilities in JMP, like Chart Builder or Fit Y by X, you realize that this is the tool for really visualizing the data and making sound decisions.”
A second benefit is the potential to better understand customers. “I told my management that JMP is the tool I wanted, because it’s how I want to do customer studies,” Sykora says. “The MaxDiff approach alone justifies it. I can actually provide things like a P value associated with MaxDiff to optimize customer studies.” Furthermore, Sykora plans to expand the use of JMP in customer studies: “I want to explore other features in data manipulation and graphing,” he says. “For example, I’d like to see what our global customers’ data might look like if graphed and overlaid on a map.”
A final advantage is the ability of JMP to actually build customer trust. “A lot of our customers, especially in the biopharmaceutical industry, also use JMP,” Kramer points out. “So when our application engineers deal directly with customers, if their data is in the same format, you’re literally speaking the same language.”
Key benefits of JMP at 3M
- Interactive visualization features in JMP facilitate fast and easy exploratory data analysis.
- A user-friendly interface makes JMP ideal for those without a background in statistics to implement more advanced methodologies by selecting from the software’s extensive menu of statistical applications.
- Robust design of experiments (DOE) capabilities, including advanced features, enable practitioners to explore and exploit multifactor opportunities.
- Support for definitive screening designs allows users to get information about main effects, curvature effects and two-way interactions at the same cost level as traditional designs.
- Quality and reliability analysis capabilities help trim excess and cut costs in testing and manufacturing processes.
- JMP helps users perform reliable validation of process improvements and faster validation of test methodologies.
- MaxDiff capabilities in JMP enable practitioners to deepen customer knowledge by gaining a more meaningful ranking of qualitative customer input.
- By using JMP, engineers gain greater customer trust in product data.
