Don't use DMAIC for Problem Solving

The problem solving template I talk about in the video is available for free through this link: www.tommentink.com/problemsol...
To check out Paul Allen's DMAIC resources and course: paul-allen.mykajabi.com/dmaic...
DMAIC is a great system for improving your processes, but don't use it for problem solving. Over 90% of operational problems can be solved much easier with the dedicated 6-step problem solving model: this will get you back to your performance standards much quicker and save you precious time and resources (which you can then spend on improving your processes with DMAIC...).
When you are dealing with a deviation from standards (not just any vague issue that somebody considers a problem), you'll most likely have an identifiable break in some machine or process caused by 1 or 2 root causes. For these kinds of process or performance problems, DMAIC doesn't give you enough structure to really help a team through the analyses and implementation steps.
It's a great framework for building better processes, but sometimes you just need to solve a problem.
#problemsolving #dmaic #continuousimprovement

[@datascience1274]

Your videos are great because they are presented in a very pratical and realistic way. I actually noticed that you don't talk about CTQ's in these scenarios. Are you saying that process speaks by itself without necessairily having to think of any CTQ in particular? The reason why I'm asking you this is that in every six sigma course they start projects with CTQ's and move forward from there, which doesn't always make sense to me.

[@TomMentink]

Thanks for that praise - I'm always very happy to hear people enjoy my work. Six Sigma enthusiasts would say that "I crave the Voice of the Customer" 🤓 As to CTQ's - they have a place, but honestly, I think that often we overcomplicate projects by forcing people to define these upfront. Do a good problem description or a 4M analysis of the basic conditions and you'll usually be fine. You will be looking at a whole number of possibly not so critical standards, but even they should be in basic conditions. The problem with looking only at Critical to Quality settings and other standards when analysing problems, is that there is usually a whole web of interconnected settings and events - bad lubrication can easily cause vibrations that interfere with your CTQ flowmeter, for instance, so investigate any scraping sounds around the line... When we are talking about improving standards beyond previously proven best results - what I call true improvement projects (in which DMAIC is a great tool, by the way), it is useful to think about which factors will have the highest impact on product quality and process stability. Because you don't want to spend time doing fully designed experiments (DOE) on every little setting you can think of. So in those cases, CTQ thinking can be really valuable.

[@datascience1274]

​​​​​​​​​@@TomMentink thank you very much Tom. When you say that you'd rather avoid DOE, do you mean that you prefer observing and collecting data to optimize CTQ with a regression model? That also makes me think of asking something else, but related to this. I wonder if you would use a similar statistical approach in lean in order to optimize some efficiency KPI. I'm asking you that because a lot of six sigma courses would talk about internal customers. To me this is often confusing because these KPI's kind of become like some sort of tolerance limits set from the internal (for instance management that set a new minimum target for cycle time, recruitment time, procurement cost ecc.). Should we improve things like these as we would with CTQ's, but considering management and departments like internal customers?

[@TomMentink]

@@datascience1274 Let me clarify that I'd avoid DOE for PROBLEM SOLVING, where we expect a root cause to our sudden decline in performance. I would use DOE for improving processes. It is definitely also possible with more basic regression models and production trials, but DOE is a great tool when seeking to find NEW optimal values for certain process settings. Here we also benefit from defining CTQ's, as we don't want to do DOE on too many factors - a DOE grows even faster than exponentially with the number of factors (it grows factorially, pun spotted 🫢). Please don't mix Specifications/Tolerances and KPI's - a specification is PRODUCT characteristic that sets outer limits on which results from a process we are allowed to count as good; a KPI is a measure of how well the PROCESS is managing to produce those products, i.e. PERFORMING. That is simplifying a bit, but the PRODUCT vs PROCESS PERFORMANCE difference is important. Mind you, there are also process standards that are neither Spec nor KPI: oven temperature, for instance - that is a setting that will most likely balance process speed, gas usage, scrap percentage; while having a fully baked bread that's not burned are the specifications (using some more 'scientific' measures to describe well baked). Cycle time is often a standard setting (the speed at which the process performs best), although you can use it as a performance indicator. Recruitment time and Procurement costs are KPI's - they are the efficiency result of trying to do both of those processes as best as you can. You cannot improve KPI's the same way as CTQ's -> CTQ's are process inputs (like setting, material parameter, specific way to operate, etc.); KPI's are not even the output of a process but the efficiency with which those outputs were created. Management is hardly an internal customer, but the production or warehouse department that receives the products you make is an internal customer (if your factory has multiple production departments within a larger flow through the whole site or to other sites within your organisation).

[@datascience1274]

@@TomMentink thank you very much, your explanation is super clear. I actually wasn't seeing KPI's like product measures, but some six sigma courses talk quite a bit about internal CTQ's, which I'm personally not a big fan of because there's the risk to mess with the whole process-product main definition. I think that effectiveness KPI's would improve when process is well within some external customer specs related to that particular KPI. In order to do so though it could be necessary to build some regression model or DOE, in the improvement phase, to manipulate the settings in a way that optimization is possible. Please feel free to correct me if I'm wrong, hoping to not take too much of your time.

[@alexeykotov6650]

Hi Tom, I totally agree with you. For me, DMAIC seems more effective where you look at the process at a whole and looking for ways how to make it better. I always use 8D approach for problem solving and quite frankly your approach is 8D for me. You just skipped the organizational steps (defining if you need to go at all or no and team congratulations) and left only steps which are related to exact problem. Btw, thank you for your problem solving template, really love it!

[@TomMentink]

Thanks Alexey. And yes, my approach is very similar to 8D. I don’t like forcing the celebration step in the model: it’s either in your culture (or program) or not, don’t put it in individual tools. And the go-no go is in each step’s sign-off with the problem owner and the way I integrate it into performance control meetings. Great to hear you’re loving the template.

[@vigneshiyer4334]

Nailed it..

[@TomMentink]

Thanks for taking the time to share that. Hope you can use the video to convince yourself and others 😉

[@zakraja2563]

Regarding services and service processes, how can Six Sigma and DMAIC show problems for service industry companies? I understand you go over problem-solving but I thought I'd ask if there are any limitations when using it for improving services.

[@TomMentink]

I would be hesitant to use the core Six Sigma statistical tools (like DOE, SPC, Cpk, R&R, etc.) in the service industry - when specifications cannot be very clearly determined and the corresponding parameters are difficult to objectively measure, the statistics tend to go out of the window. You can use many of the support tools linked to Six Sigma, including DMAIC, which is an all-round good improvement method, also for service industries. I don't think the service industry faces specific limitations regarding DMAIC. I do think that DMAIC is very well suited for focussed improvement, but not the easiest root cause problem solving tool to get back to standards (the core message of this video).

[@loveforever5687]

As a Quality Engineer I choose and pick what analysis tool or method works for the current problem and go with it. It's always worked. I dont have a degree, but i have 8+ years of experience as a QA tech and worked my way to an Engineer. those 8 years of experience includes everything from a QA inspector to Engineering and everything in between, including systems, Financial and customer service. I have no problem going head to head with some of the best engineers in the world in my field. Why? because i've done it all, I know the industry standard better than any engineer that i know of and have implemented better solutions than any engineer that i worked with. Why, because i'v done the raw and dirty works that engineers dont usually do or want to do.

[@TomMentink]

With such a background, you have collected the tools and experience to size up problems when you see them. So you're right not to limit your choices up front. But if you want to make correcting deviations from process/performance standards (aka problem solving) a common good in your organisation, having a structured and relatively easy-to-use method in place will greatly benefit you. (and it might also free up your time, so that you have the resources to do process improvements, where you push the performance of processes beyond their current standards)

[@mariejane1567]

I'm sorry this is splitting hairs. The DMAIC approach is all about scale and speed. Which step would you not do conceptually? Maybe it comes down to training and application and user experience. The DMAIC approach lines up with the ISO 9001 standard for corrective action. This is another case of hammer and nail......

[@TomMentink]

Hi Marie, I will admit that this video and it’s title were designed to draw attention, and yes, DMAIC can be used for problem solving and process improvement if you know how to apply it correctly. I do see that for many companies the distinction between restoring standards (root cause problem solving) and improving standards (improvement team) is not clear enough, even though it will help you a lot in practice. That’s why I have good results implementing two methods (with different names) for problem solving vs process improvement. With that in mind, DMAIC doesn’t give much support for root cause problem solving - yes, you can definitely use it for it, but the very directed structure of “team+mission, define problem, temporary quick fix, find root causes, implement countermeasures, sustain the corrections made so that the problem doesn’t return” does help teams explicitly think about those steps that have the highest likeliness of success. DMAIC is a general improvement method, which requires the user to really know what they’re doing. My favorite 6-step is more of a ‘guide-by-the-hand’ tool whose wording and templates are fully geared towards expecting there to be a root cause to the problem you’re dealing with and guiding a team without much CI experience to find and solve that cause together. Hair splitting, maybe a bit, but there is method to this madness (and better results in my own practical experience).