Below, you’ll find a transcript of our Q & A with Matt Hansen, founder at www.statstuff.com. If you’re into lean, six-sigma, continuous improvement, or jellybeans, Matt’s website is a completely free resource, and has a bunch of free training and materials. We caught up with him last week and here’s what Matt had to say about the “irony of the measure phase of the DMAIC cycle”, and why prizes for jellybean counting competitions might help.
Before you start reading, consider one of Matt’s key questions, as you consider how you measure operations and projects.
If I had a huge jar of jellybeans, and you had to guess how many red ones were in the jar, would you measure them precisely? Or, would you take a sample? Is your answer different if your reward for the right guess is a T-shirt? What if the reward is $1,000,000? How might you want to measure now?
Brent: We’re interviewing Matt today because we really wanted to talk to him about an active post he put up on Linkedin in a Continuous Improvement and Lean and Six-Sigma group we both belong to. Essentially, Matt put a post up about the measure phase of the DMAIC cycle, but it was an active discussion and a really relevant topic for us here at SITEFLO. Welcome Matt.
Matt: Thanks for having me Brent.
Brent: Can you give me some perspective on DMAIC for the unitiaited when it comes to lean and six-sigma, and what it means?
Matt: Sure. DMAIC is 5 different phases in a process. It starts off with the define phase, which is when we try to understand a problem that we’re trying to solve. And then we move to measure phase, which is when we try to gather reliable information around the problem we’re trying to solve. After that we move to the analyze phase, which is when we’re trying to apply statistical tools and analyses to see the root cause based on the analysis of the data we gathered on the measure phase. Once we understand the root cause, we can try to figure out how we might fix it, which is when we move to the improve phase to figure out what things we can do and implement to fix the root causes. After implementation, we move on to the control phase, because we don’t want the original problem to rear its ugly head again. So, we put certain controls in place, and those controls are around solutions to prevent that problem from happening again, so we can sustain solutions. I describe the DMAIC flow as a scientific method that’s adapted to business, and It’s common sense, but it has a methodical flow.
Brent: In the Linkedin group there was a lot of discussion about tactics, and what can happen when you get to the measurement phase of the DMAIC cycle. Can you talk a bit about what tactics people employ to get good measurements?
Matt: The tactics you might use really vary from organization to organization. So, it might vary from using existing systems that they have embedded within the organization, or they might try to create new methods to collect data if they don’t have an existing system to collect information, they might have to create a manual method. And then they have to devise some way of how they’re going to gather that information. Often, we see a lot of people walking around with clipboards, or maybe with a laptop, just observing what’s going on through the process and documenting critical points that we think we want to track. We gather that compiled information to work through and analyze the data. So, it really varies. The ideal method is when we have a system in place that’s already been proven and tested, where you have reliable information at critical points within the process. Like, if you have time stamps, or other kinds of things that factor in the movement and flow of products or activities. It’s not always easy, because sometimes the problem you’re trying to fix doesn’t have a measurement, so you have to create those measurements. That’s when it gets really challenging. And when we work through those challenges and we get the data, just having the data isn’t enough. My experience has been that you have to trust your data, so you might have to employ some additional effort to make sure this is data you can trust that comes from a trusted source. You need to know who recorded the data, or the time they recorded it. When you’re dealing with multiple shifts across different days, you may have people who recorded their data only within one shift because that’s what most convenient. But, that may not be representative of the entire process. So, you have to take a lot of these factors into consideration. You really have to move on to the measurement system analysis, or MSA, where you test for the accuracy, reliability, and reproducibility of your data. By working through several tests as part of that MSA proces, you can walk away from the measure phase knowing whether or not you can trust the data. And then you move onto the analysis phase. If I can’t trust the data? What’s the point?
Brent: That really gets at the heart of some of the stuff we run into everyday, which has to do with trying to validate whether or not the data that people have is trustworthy at all. It’s really tough to get at that. With modern technology, there are so many ways that we can collect better data, I think there are a lot of solutions out there for people to really nail that measure phase.
So Matt, why do organizations typcally fall short on the measure phase of this cycle in the first place?
Matt: Well, this is really relevant to the question that even prompted that discussion on Linkedin. It’s easy to blame it on impatience or laziness, but I think it’s more than that because I don’t think anyone would do that intentionally if they understood the risks that they’re taking. What it comes down to is an assumption that people are making about risks, and if they really understood the assumptions that they were making and the benefits of the process they’re working through, they wouldn’t fail at the measure phase in the first place. The way I described it in the Linkedin group was about the irony of the measure phase, and people who work through it quickly. So, they look for quick ways to get the data, with the intention of getting to a fast solution that they can implement. What I find is, as they complete the define and measure phase, they get to analyze and realize they don’t have the right data, and then they find themselves having to go back to the measure phase. And again, I think it goes back to the assumptions they’re making. Take my jellybean example. Imagine you had a huge jar of jellybeans in front of you, and you had to guess how many red jellybeans there were in the jar. Well, there are two approaches you can take. One, you can pour out all the jellybeans, separate them into colours, and count them one by one. It will take a lot of time, but you’ll be very accurate in your approach. The second method is if you wanted to scoop out a sample, and extrapolate and estimate to figure out how many red ones you may have in the entire jar. Both approaches are valid, and that second example is faster, but it’s not as accurate. We use the second method a lot in statistics, so we don’t have to sample an entire population. The problem becomes when you’re trying to understand the risks and the benefits. If the prize is a T-shirt after you count the jellybeans? Well, no big deal. You don’t want to spend a lot of time on it. But, what if the prize was $1,000,000? Well, nobody want to guess anymore and we want to be really precise about those measurements. I’m going to count it because I need to be accurate. If we really understood the risks of bad data as we move through the DMAIC flow, we would understand that it’s worth getting data that you trust so you can move on. It gives us confidence that we’re getting to the right solution after all.
Brent: Before we leave, I wanted you to tell people a little bit about Statstuff. Can you tell people a bit about what you’re doing, and why they might want to visit the site?
Matt: Sure. So, I boast it as the only free online source for the complete training content for lean six-sigma. If you go to any other training source, at your local college or online, people end up hundreds if not thousands of dollars to go through their training. They usually dedicate several days and weeks to go through the entire training, but everything I’ve got on Statstuff is the same kind of content they can get from any other training organization. Anyone can access the content for free. A lot of the videos are free and public, and some require registration, but, it’s still all free. I’m doing this because this is a lot of math and common sense information, and I learned and benefited from a lot of other sources. I just don’t feel right about holding this back from people so I just offer it, and allow people to use the information. Statstuff, from what I understand, has one of the largest compilations of comparisons between other training organizations. It’s really hard to compare costs, time investments, and quality between training organizations. I offer a list so you can compare their pricing, requirements for certification, and all of those wonderful things you want to know about.
Brent: There you have it. Completely free resource. Generally, we don’t do a whole lot of advertorial stuff but this is entirely free and it’s been helpful to us. Hope it’s helpful for some of our readers.