A transformative training program teaches applied statistics – and the tools to go with it
At leading vaccine manufacturer KM Biologics, an innovative in-house statistical literacy program goes beyond theory to teach practical data skills to scientists and engineers. This approach is already bearing fruit as cost-saving improvements come online.
|Although KM Biologics had for years offered traditional industrial statistics education to scientists and engineers, many participants felt there was a disconnect between statistics in theory and in practice. “Even though we were given the opportunity to learn basic statistical knowledge, it was difficult for most participants to acquire sufficient understanding to actually use the skills they were learning,” explains Shinji Tochihara, a commissioned specialist in product development at KM Biologics. “That’s why I searched for a better way to improve the statistical literacy of our employees.”
|To make KM Biologics’ statistical education offerings more applicable to domain experts’ day-to-day work in R&D and manufacturing, Tochihara adapted a free online statistics program called Statistical Thinking for Industrial Problem Solving (STIPS). The course, developed by JMP Statistical Discovery, provides roughly 30 hours of material specially designed for a science and engineering audience. Tochihara tailored STIPS learning modules to suit KM Biologics’ unique needs by pairing the self-paced online offering with scheduled in-house trainings so that participants could gather to develop a community of statistical practice.
|The company’s reimagining of its statistical education offering has already paid off by adding immense value via improvement opportunities. In one instance, a training participant used what she learned in STIPS to build control charts in JMP that pointed to an improvement with dramatic cost-saving potential.
Leading global vaccine supplier KM Biologics first gained recognition for its inactivated smallpox vaccine over a half-century ago and has since developed products to vaccinate against everything from typhoid to the novel coronavirus, Covid-19. Now a subsidiary of Meiji Holdings, the Kumamoto, Japan-based company distributes more than 15 million vials of vaccine annually. In the past four years alone, KM Biologics has had 14 new or improved vaccines in development, thanks in large part to a robust R&D organization of more than 160 of Japan’s most elite scientists and engineers.
As biologics development – like many technical fields – has become increasingly data-driven, leaders at KM Biologics recognized a need to upskill domain experts in statistical best practices. "While statistics study sessions have been held here at KM Biologics for quite some time, many scientists were still struggling to fully understand how to apply statistical methods to their work in meaningful ways,” explains product development specialist Shinji Tochihara. “Even if participants left a session thinking they had grasped a concept, putting that concept into practice was another matter.”
After an audit of existing training offerings revealed a disconnect between statistical concepts and participants’ ability to apply those concepts to on-the-job challenges, Tochihara began building a more effective in-house statistical education program that would focus on helping domain experts cultivate the necessary skill sets. He now leads a voluntary training program that is attracting attention for its success in raising the statistical literacy of the entire organization.
“If we can grow the number of people who understand even basic statistics, then we lay the foundation for those people to continue studying more advanced concepts on their own – and to help their colleagues develop greater statistical literacy as well,” he says. It’s all part of an enterprisewide capacity building process. “Ideally, the learning that goes on during a workshop should not stop there,” he adds. “The skills participants develop should be useful in their actual day-to-day work.”
Going beyond classical coursework to help engineers and scientists develop a practical skill set
A longtime user of industry standard software JMP®, Tochihara had previously come across JMP educational resources; however, KM Biologics had yet to implement them in any formalized way. Of particular interest was Statistical Thinking for Industrial Problem Solving (STIPS), an online course offered for free by JMP to anyone curious about learning new ways of solving problems with data. STIPS, Tochihara says, was especially appealing in that it pairs theoretical skills with useable, practical skill sets – and it links beautifully with the many functionalities in JMP software. “Dynamic animation in STIPS really helps trainees understand,” he adds.
Developed by experts at JMP with extensive industry experience, STIPS is oriented toward problem-solving applications commonly used in business scenarios, ranging from R&D and manufacturing to marketing and consumer research. The course offers material for users of all levels, from statistical novices to advanced practitioners looking to expand their knowledge. Comprised of seven, self-paced modules that can be taken in any order, STIPS offers roughly 30 hours of instructional videos, demos, questions and exercises – all of which can be tailored to suit individual preferences.
“I’ve found that the flow of the course material is appropriate for both beginners and more advanced students,” Tochihara attests. While at first he used STIPS as a supplement to existing materials, such as textbooks, as the workshop progressed, he realized that supplemental materials were unnecessary.
“We then decided to develop a workshop entirely based on JMP and STIPS,” he says; now STIPS modules are supplemented only by accompanying PowerPoint slides. “We also make use of the exercises and quizzes that measure the degree of achievement – and have devised other ways to make the course easier for students to use.”
By incentivizing colleagues to complete STIPS coursework in the classroom setting, Tochihara explains, he is not only upskilling colleagues in statistics; he is also giving them a tool to practice what they have learned. Furthermore, the training has created a community of statistical practitioners within KM Biologics where colleagues can work together to solve data challenges.
Statistical upskilling leads to added value for the business
What originally began with just one student – Tomoko Sakazawa – in a single workshop has since grown to five training options geared to different skill levels and divisions. In 2022, a 22-student cohort comprised 12 people from R&D and 10 from production. Graduates have made extensive use of control charts and design of experiments, among other things, and, notes Tochihara, "in the future, we will use methods like partitioning and regression more frequently to optimize the way we use process data."
One workshop participant from the manufacturing department of KM Biologics’ Kikuchi factory, Tomoko Sakazawa, says, "STIPS is excellent in that you can learn statistics in a way that is in line with the actual work of the manufacturing department. We use what we’ve learned from the course in our everyday work, especially when creating control charts with JMP."
For example, last year Sakazawa created a control chart with JMP that used process data from the product line she works on, which allowed her to observe a strange and unanticipated signal. She brought her findings to on-site colleagues who, after seeing the data, made adjustments to the line that led to a monumental cost reduction. Sakazawa was later recognized with KM Biologics’ presidential award for this valuable and ingenious improvement.
"Since I started creating control charts with JMP, I’ve been able to dramatically reduce the man-hours of work for specific tasks,” she says. “Before JMP, I was building control charts in Microsoft Excel, which was time-consuming and error-prone because I had to spot any abnormality in the data by eye. With JMP, I can so much more easily use Western Electric rules [an approach to detecting out-of-control conditions in statistical process control] to obtain results – saving time and improving accuracy in the process.”
Program graduates have also implemented more advanced design of experiments (DOE) approaches to streamline experimentation, cutting costs and time in the process. "I was doing DOE in JMP in my department before [joining the course] and although I had some experience, I still lacked a deep understanding of the data and how to interpret the results obtained by DOE,” reflects research scientist Tatsufumi Onchi. He says that since his participation in one of the STIPS-based workshops, his understanding has greatly expanded: “Now I’m proceeding with DOE and have recently been developing variable effect models like split experiments and mixed models with Mr. Tochihara’s help.”
As for Tochihara, he has no doubt that efforts to build in-house statistical literacy are bearing fruit. "Even as an instructor, I never stop learning. There is always so much more to discover about statistics,” he says. “I would urge the next generation of scientists and engineers not to give up – and to learn by doing. As long as I continue to teach, I will strive to make sure that anyone who works with data has a basic knowledge of statistics. Ultimately, my hope is that statistical education will take its place as a cornerstone policy at our company."