Minitab’s acquisition of Effex (now Minitab DOE by Effex) brings advanced experimental design capabilities into a broader analytics ecosystem, making it easier for organizations to design smarter experiments, reduce costs, and accelerate innovation.
To explore what this means in practice, we spoke with Professor Peter Goos, co-founder of Effex and a leading expert in statistical design and analysis of experiments. In this conversation, he explains the fundamentals of Design of Experiments (DOE), why it remains underutilized, and how modern approaches are changing the way organizations approach experimentation.
How would you describe Design of Experiments to people who are new to it?
Design of Experiments (DOE) is a systematic approach for determining how input factors influence output factors. Often, when studying a system, you want to understand what happens if you change or adjust a specific input. DOE provides a structured method for examining the impact of all the inputs, or "buttons," in your system. Essentially, DOE helps you study the effect of all your inputs on the system's outputs.
Could you give a practical example to make it more tangible?
If you want to bake bread, you need to consider several inputs. How much flour will you use? How much water? What baking temperature will you set? Changing the baking temperature, the amount of flour, and/or the amount of water will influence the taste, consistency, and volume of the bread. DOE offers a systematic approach to link the different levels of the different inputs to the taste, consistency and volume. That link takes the form of a statistical model, which can predict the output for any combination of inputs you choose.
How did you end up specializing in statistics, particularly in Design of Experiments?
Long story short, I would say: “coincidence.” I was studying business engineering, and during my studies, I only briefly encountered Design of Experiments in a Quality Management course. At that time, DOE didn't immediately resonate with me, but later, during my PhD, the professor who hired me, Prof. Martina Vandebroek, was very interested in DOE. She pushed me in that direction, and I soon grew to love the field, especially after working with a consultant on practical applications and collaborating with various companies in the automotive sector.. I also appreciated the optimization aspect, which was my first love. I had originally wanted to pursue a PhD in optimization. Moreover, the practical applications and impact on companies were rewarding. Seeing their satisfaction when faced with good results motivated me to continue in this field.
What's the coolest problem you've solved using Design of Experiments?
One of the more interesting and recent examples is when we helped Kellogg's improve their recipe for Pringles potato chips. Over the years, the production costs had increased, and the fat content was also a bit high. Kellogg's wanted to develop a new recipe that would reduce both the fat content and production costs. They conducted a large experiment, testing various new ingredients. Using Design of Experiments, we analyzed and modeled the data, and then performed optimization based on the statistical model. As a result, they were able to significantly lower both the fat content and the production cost. It was one of the most impressive and rewarding cases we've worked on recently.
Let’s discuss Design of Experiments. How DOEs it compare to other methods? For example, why did Kellogg's choose DOE instead of simply baking chips and checking the results?
Many engineers prefer a classical 'one factor at a time' approach, but Kellogg's needed to use Design of Experiments. Increasing the amount of one ingredient inevitably required reducing the amount of another ingredient, which is why the traditional 'one factor at a time' was infeasible. At Kellogg's, changing one input affected others, so they had to consider all inputs simultaneously. This key feature sets Design of Experiments apart from the 'one factor at a time' approach. It may seem counterintuitive because if you change multiple factors at once and see an effect, whether positive, negative, beneficial, or detrimental, you don’t immediately know which input caused it. That’s where statistics comes into play. The statistical technique, multiple linear regression, helps to disentangle the effects of all inputs. The only requirement is that you change multiple inputs simultaneously, according to a well-designed DOE. Once you collect the data, multiple linear regression analysis identifies the specific contribution of each input.
When should someone consider using Design of Experiments? For example, if I were a scientist or engineer, when would you say, 'Now is the time to use DOE’? Can you give an example?
I've given this a lot of thought. When you have two or more inputs, you should consider Design of Experiments. With just one input, trial and error can work, but as soon as two factors are involved, subject matter expertise alone often isn't enough. When two inputs interact, things get complicated. We call the joint impact of two inputs, other than the sum of their individual effects, an interaction effect. Interaction can be synergistic or antagonistic. Oftentimes, they are hard to assess or predict with subject matter expertise alone. In such cases, an empirical approach is needed to test multiple inputs simultaneously and quantify any interaction effects. That’s where DOE comes in. It provides an efficient way to identify important interactions between multiple inputs. Instead of endlessly changing variables, DOE helps manage the process effectively whenever you have two or more factors.
So, are you saying that using Design of Experiments is also more cost-efficient?
Definitely. Many people still believe that Design of Experiments is inefficient because they think it involves testing all possible combinations of inputs. But that's not true. Modern, 21st-century Design of Experiments is no longer about testing every possible combination. Quite the opposite; it’s about strategically selecting which combinations to test while considering the budget, even if it's limited.
How important is statistics, and by extension, Design of Experiments, in the innovation process?
During the innovation process, there’s a lot of uncertainty since you know little about the new product or process and have limited data. To make data-driven decisions, you need to collect the right data. This is where DOE excels. It’s highly effective for gathering high-quality data economically, even under time pressure or in costly environments. In uncertain situations, small but powerful experiments are key, and DOE provides the tools to run them efficiently. Proper statistical analysis then allows you to distinguish the signals from the noise—specifically, the true effects of the inputs on the output versus the inevitable random variation in the data.
Why is Design of Experiments still relatively unknown today? And why Doesn't it receive more attention in colleges?
There are several reasons for this. First, DOE isn’t taught enough at universities. Even when it is, students often lack the real-world experience to see its value. DOE is better appreciated by those who have encountered the limitations of relying solely on subject matter expertise and understand the need for systematic data collection.
In companies, who should be aware of Design of Experiments?
Anyone in engineering, R&D, or production focused on continuous improvement should be aware of DOE. It’s also important for management, including C-level executives, to have some knowledge about DOE so they can recognize its potential for optimization and innovation.
What would you say is one of the biggest trends or evolutions in DOE?
The biggest evolution is the rise of optimal design of experiments, which allows experiments to be tailored to specific constraints like budget, feasibility, and time. This makes DOE far more flexible and cost-effective than earlier approaches.
Do you think there's still room for further evolution? What do you see as the future of DOE?
A major shift is moving from a two-stage experimentation process to a single, more efficient experiment. New approaches like definitive screening designs (DSDs) and OMARS designs combine screening and optimization into one, allowing companies to move faster and innovate more efficiently.
With the rise of AI, is DOE still necessary?
There will always be a need for DOE. AI can help analyze historical data and suggest paths forward, but when exploring new combinations where no data exists, experimentation is essential. DOE and AI should work together.
Why is there a need for new DOE software like Effex?
It’s critical that advanced, cost-efficient experimental designs are widely accessible. Effex provides access to cutting-edge designs like OMARS and uses a cloud-based platform to continuously improve the quality of experiments available to users.
What advice would you give to organizations that have had negative experiences with DOE?
Failures are usually not due to the method itself but to how it’s applied. Choosing the right inputs and having sufficient subject matter expertise is critical. Without that, any approach—not just DOE—can fail.
What’s the minimum knowledge required to start using DOE?
Two things are essential: understanding multiple linear regression and having the right software to design and explore experiments effectively.
Which industries are underutilizing DOE today?
Food companies, in particular, could benefit more from DOE but often face challenges due to variability in biological materials, which makes experimentation more complex.
Is “doing nothing” really the biggest competitor to DOE?
In many cases, yes. Some organizations are satisfied with current processes, but that approach isn’t sustainable long term. To remain competitive, companies will need to innovate, and DOE plays a key role in that.
What’s the most compelling way to convince leadership to adopt DOE?
Case studies and real-world success stories are the most effective. Seeing tangible results helps demonstrate the value clearly.
Design of Experiments is not just a statistical technique. It’s a practical, scalable way to make better decisions in complex systems.
With Minitab and Effex now combined, organizations have access to more advanced experimental design capabilities, including modern approaches like OMARS designs, within a connected analytics environment.
If you’re looking to reduce cost, accelerate development, or make more confident decisions, the team at Minitab is here to help you apply these methods in your own work.