Manager of Global Growth Analytics, Hexion
User Reference Manager, JMP
Dan Fortune is Manager of Global Growth Analytics at Hexion Inc., a specialty chemical company driving innovation in thermoset resins. A Lean Six Sigma Master Black Belt, Dan joined the company five years ago to provide statistical modeling and analytical support to R&D activities in resins for engineered wood and slow-release nitrogen fertilizers.
Dan previously served in a range of leadership roles in business excellence and quality engineering for TechnipFMC, Mitsubishi Caterpillar, Maytag and Texas Instruments. He holds a Master’s in Quality, Reliability and Statistical Engineering from Arizona State University.
JMP User Reference Manager Meg Hermes spoke with Dan during JMP Discovery Summit at SAS Headquarters in North Carolina.
Meg: I understand you’re the only person without a chemistry background in Hexion’s R&D group! How has that influenced your perspective on the ideal balance between chemistry and statistical expertise in industry?
Dan: At Hexion, we are evolving in the right direction as a company: increasing statistical contributions to R&D without reducing the theoretical chemistry contribution.
In general, the human brain has a difficult time understanding and applying the concept of complex chemical interactions. And when you throw in nonlinearity, there’s just no way to do it effectively without statistical tools! Sure, it might be easy for chemists to experiment in incremental amounts, [looking at one factor at a time], but it's not an efficient way of doing experiments. When you've got a high-dimensional design space, you need [design of experiments (DOE)] to help determine whether you've sufficiently explored it.
And that’s what I help with: I help chemists gain new knowledge – and I provide support when they’re exploring the design space. I'm working to change [how chemists think about experimentation], and not only does statistics add significantly to their knowledge and result in better understanding of the design space, but it also reduces the total number of experiments they have to run.
While I don't do formulations myself per se, I do make recommendations based on what the software lets me know about the way [factors] fit together. I think you need a reasonable theoretical background to convince chemists (or even other scientists) that you can help them to be even more successful by narrowing the window of experimentation, reducing bench time and accelerating R&D. We have found that when utilizing the software, the predictions are sometimes outside industry norms and assumptions, which has been surprising to me and the chemists…and they are yielding great results in the lab and in field trials.
That said, to date, my recommendations have never once been rejected by a chemist. So yes, the fact that PhD chemists take my advice on chemical formulations is pretty interesting. And the more we’ve learned, the more we’ve accelerated our development process in a way that has brought about some fairly substantial advances in our products.
Meg: How have statistical approaches like DOE been received?
Dan: When introducing a new concept, you're always going to have some people who are ready to jump on board right away and others who are reluctant – it’s human nature. But nothing sells like success. The more that our executives saw the types of improvements that we were making relative to our products, the more they insisted upon [DOE] being used in new projects. You always want to have pull from your customers – who, in my case, are chemists – but sometimes the executives push too.
Meg: I understand you’ve gotten the green light to hire someone in a new role providing statistical support in R&D. How did that come about?
Dan: To start with, we don’t plan to add headcount without seeing a reduction elsewhere – we’ll use attrition to enable a new position. So while it may not happen immediately, it means I've got time to carefully consider the types of qualifications that I want for the position. My guess is that it will be a marriage of statistics and computer science. Over the summer I had an intern – a computer science PhD candidate – and I got a glimpse at what at least that one computer scientist brought to the game. We were able to do a lot.
The other question is which tool they will use. I had just about decided that I wanted to hire somebody who was an R guru, but [my intern] had been working in Python for a long time and I was impressed with what he was doing with his programming. We decided to do a comparative review of neural network applications in R, Python and JMP. And JMP's was really, really good. He started switching from Python to JMP to do quick data manipulation, table manipulation – things that are much more difficult to do when you have to type them out rather than just click a couple buttons in JMP, and you're instantly there. It was enlightening. And you can always do that integration from Python and R back into JMP to enable any capabilities that might not be currently present in JMP.
It’s really important to say that people who are getting into this field should develop a strong skill set in JMP or R or Python – the more the better. Being able to move around one of those tools with ease and no hesitation is a super valuable skill. So that's one thing I'm certainly going to look at in any job candidate.
Meg: Speaking of advice, what would you suggest to chemists who are looking to learn JMP?
Dan: Chemists do everything in Excel. So I would say to them: switch to JMP for one month. Do the 30-day free trial, and then make your decision about whether you're going to continue. Give it an honest 30 days. I did it myself – 30 days, no Excel – and I've never switched back. That experiment increased my capability and skill level in JMP tremendously.
The 30-day JMP challenge – that's an easy one. People are so accustomed to using Excel, they won’t learn another tool's capabilities. But once you learn JMP's capabilities, my guess is that you will never use Excel for data manipulation ever again.
Meg: Start learning to use Excel in middle school and by the time you have a PhD, 15 or more years later, Excel is just the default…
Dan: Right. And learning JMP doesn't mean you have to stop using Excel. It's just that you will find that there are times when you really need to think of data from an analytics perspective rather than from just a presentation perspective.
Using JMP will cause you to start thinking in terms of the way that your data is analyzed. For example, one row is one instance. Data is not analyzed by looking at one section of a window at a time. It may have all these colors and spaces and everything's separated nicely to where it prints out on a sheet of paper that looks pretty. But that's not data analysis. Thinking of data from the perspective of how it is analyzed will help you significantly.
Meg: Aside from in statistics and computer science, most graduate and undergraduate students use Excel in their coursework – even though industry is rapidly evolving toward more dynamic tools. From the vantage point of someone who is looking to recruit and hire new technical talent, how do you think universities could better prepare students for industry careers?
Dan: I'll tell you, my advice to faculty is simple: You need to get people off Minitab. Minitab is a worksheet and its capabilities are super low. Is it better than just an add-in for Excel? Maybe. But I think there are a lot of folks who learn Minitab in college who would be better off learning JMP.
It's interesting to see textbooks that have Minitab examples up until a point, and then the rest of the book is JMP. So when you go to the tougher stuff, you go to JMP, and then you use Minitab for the easy stuff? Why even do that? Why would you waste your time with Minitab?
Meg: We’re having this conversation during Discovery Summit, and I know you've been here at JMP’s offices in the past – including when you sat on a customer steering committee. In other words, you’re really engaged. What value does that level of engagement deliver both to you professionally and to Hexion?
Dan: When you come here [to Discovery Summit] it’s all about those little tips and tricks that are the key to efficiency. Every time I’m around someone who uses JMP – even if they’re not expert JMP users – I learn things because they know a trick that I don't know.
As you learn those tricks, you start cruising through JMP. And as you learn more techniques and which techniques work in certain types of situations, you're able to better address real world problems. It becomes a snowball rolling down the hill; your knowledge and capabilities are advancing. As a new user, you always start slow. It's how you advance your knowledge to get the snowball rolling with its own momentum that makes the difference. Being around JMP people really helps accelerate things.
You always come out [of Discovery Summit] with new ideas. You've just seen people who are very excited about what they're doing. You can’t help but to get a little boost in energy and motivation.
And you pick up things here that you hadn't thought of before. I can tell you that I have seen some super clever things at Discovery Summit. For me, clever is getting some incredible result by something very simple. I've seen that here, and it was just like, "Well, that is super simple, but wow, is that effective." I can apply it maybe not exactly like they did, but there's an application to my work.
For example, one time, I brought back a technique that came from the world of autonomous vehicles. I saw that it could work for chemistry. And it has been super effective for me ever since.
People start congratulating you on your success because you started using a technique you learned at Discovery, and it's a pretty powerful statement relative to the effectiveness of attending these conferences.