Mistake-proofing expert Martin Hinckley joins the podcast to explain why traditional inspection methods can't catch rare mistakes — and what it actually takes to eliminate them. We cover the difference between mistakes and variation, the real meaning of poka-yoke, and why good mistake-proofing always makes the work easier, not harder.
This is LeanBlog Podcast episode #42 with Dr.C. Martin Hinckley, of the firm Assured Quality. He is the author of the book Make No Mistake!: An Outcome-Based Approach to Mistake-Proofing, available through Productivity Press. We'll talk about his book and approaches for teaching people how to develop mistake-proofing in processes.
For earlier episodes, visit the main Podcast page, which includes information on how to subscribe via RSS or via Apple Podcasts.
LeanBlog Podcast #42 Key Points & Links
- Some mistake proofing examples from Dr. Hinckley's site
- Upcoming webinar on mistake proofing on May 15 (via http://www.visualworkplace.com/)
If you have feedback on the podcast, or any questions for me or my guests, you can email me at leanpodcast@gmail.com or you can call and leave a voicemail by calling the “Lean Line” at (817) 372-5682 or contact me via Skype id “mgraban”. Please give your location and your first name. Any comments (email or voicemail) might be used in follow ups to the podcast.
Transcript:
Mark Graban: Hi, this is Mark Graban. You're listening to episode number 42 of the Lean Blog podcast for May 7th, 2008. Our guest today is Dr. Martin Hinckley of the firm Assured Quality, and we'll be talking about his book entitled “Make No Mistake: An Outcome-Based Approach to Mistake-Proofing,” from Productivity Press.
Dr. Hinckley works quite a bit with a good friend of the podcast, Gwendolyn Galsworth, who will be featured once again in a new episode of the podcast in upcoming weeks. So I hope you will come back for that and other future episodes. Additionally, if you're interested in a webinar by Dr. Hinckley coming up on May 15th, 2008, you can register for that via Gwendolyn's website at www.visualworkplace.com.
As always, thanks for listening.
My guest today on the Lean Blog podcast is Martin Hinckley. Thanks for joining us.
Martin Hinckley: My pleasure, Mark. I appreciate the invitation.
Mark Graban: I was wondering if you could start off by introducing yourself to our audience — tell us about your background and your experience with Lean, and also introduce your book that we're going to be talking about today.
Background and the Link Between Complexity and Defects
Martin Hinckley: Sure. For about 25 years, I worked at Sandia National Laboratories and was the lead engineer on some major Department of Energy projects. I went back to school at Stanford, sponsored by Sandia, and while there I was working with Professor Phil Barkin, who is a leader in design for manufacturability.
We were looking for the link between product design and overall quality, and found that we couldn't find a single company that had been able to roll up statistics to predict defect rates in their products. From that, we started to look for the things that would really make a difference in quality. We found a strong link between the complexity of the product and the defect rates. The only conclusion we could reach was that mistakes are really the driver behind most quality problems today. That led to the book I wrote, “Make No Mistake,” which describes techniques and methods for mistake-proofing.
Mistakes vs. Variation
Mark Graban: Talking about complexity, mistakes, and variation — how would you describe the difference between mistakes and variation?
Martin Hinckley: The most significant change in quality thinking came when Henry Ford developed standard gauges, which allowed us to detect differences in one product from another. Then in the 1920s, they identified variation as a cause of quality problems.
In every process, there's a distribution in how a task is performed. If I take a hole-drilling process, for example, I can gauge the diameter of the hole and determine that sometimes it's smaller or larger than the nominal — normally a bell-shaped curve.
The problem is that a mistake is a probabilistic event rather than a statistical one. It occurs rarely. I may occasionally forget to drill a hole, use the wrong drill bit, drill in the wrong place, or drill partway through instead of all the way. These events occur — undetected — about once in every 10,000 to once in every 100,000 operations. They're rare, but there are many types of mistakes that can occur. And as our quality has improved and the problems from variation have decreased, mistakes have become the most dominant source of quality problems.
A study by GE Aircraft Engines looked at all the problems they had to respond to in the field that were not scheduled maintenance. Over a multi-year study, they concluded that all but one problem was traceable to mistakes. Only one could be traced to variation.
Mark Graban: Is it fair to say that when mistakes are so rare, inspection becomes an ineffective method for detecting them? People come to expect that the mistake won't be made, and they get lax about actually catching it when it does happen.
Martin Hinckley: It's almost impossible to characterize mistakes through traditional inspection methods. Because mistakes are so rare, if you're doing sampling inspections of, say, one in a hundred parts, you would have to sample a hundred million operations before you could characterize the frequency of one kind of mistake. Most of our statistical methods simply aren't giving us good information about mistakes. Even if we do all kinds of traditional sampling inspections, we can never control mistakes that way. That's why mistake-proofing requires a totally different approach — it requires 100% checking.
The reason we moved to sampling inspection was to save money, and the key is that effective mistake-proofing can actually be more cost-effective than traditional inspection. To illustrate: in a hole-drilling process, the traditional approach is to drill the bolt hole pattern, take it to inspection, have them gauge the hole diameters, and maybe two days later get a report back — after you've already made a bunch of errors.
With mistake-proofing, I would have a sensor that watches the drill plunge through the part to make sure it went through the right number of times. I couldn't start the machine until everything was in the right location. It would count the number of drilling operations, and I couldn't remove the part until it was done. There would also be a counter tracking how many times the drill has been used, so I have to replace it before it produces an out-of-tolerance part.
Then I take the part off the machine — and it's already inspected. No downstream inspection of that process. I know it's right, and I can execute the process faster because I'm not rechecking all my work.
Overcoming Objections to Mistake-Proofing
Mark Graban: Do people ever hesitate when you say error-proofing can be less expensive than inspection — especially if they're already doing inspection? Do you hear things like, “It's okay, I have really skilled, careful people”? How do you respond to that?
Martin Hinckley: A lot of people have that kind of emotional response, but the data doesn't support it. The problem is that all people — no matter how careful — make mistakes. Studies done by the Department of Energy show that error rates between the best and worst of us are generally about a factor of two. They're not huge differences. We all make mistakes.
It's almost impossible to control mistakes without systematic countermeasures. We know of companies that have delivered a million products with a single defect — with mistake-proofing in place.
What's interesting is that properly implemented mistake-proofing makes the process easier, not more difficult. A counterexample: in some hospitals, efforts to mistake-proof medication delivery have been done poorly and created near-disasters. Nurses use medication carts that must be attended at all times. They scan the patient's barcode and then the medicine's barcode. But if an emergency happens in another room, they can't leave a cart with unlocked medications. So a lot of times nurses circumvent the process by scanning a barcode in a book rather than on the patient.
That's a case where the implementation hasn't made the task easier. Good mistake-proofing always makes the task easier.
Mark Graban: Right — so it doesn't create the incentive to work around the process.
Martin Hinckley: Exactly.
Poka-Yoke: What the Word Actually Means
Mark Graban: I wanted to ask about the phrase “poka-yoke” — what's the correct pronunciation, and what does it really mean?
Martin Hinckley: It's definitely Japanese, and it's normally translated as “mistake-proofing.” The verb comes from “yokeru.” I lived in Japan for a couple of years. The common mistranslation comes from the fact that Japanese has many different levels of language — honorific forms, casual forms, and command forms. The “yoke” form — the jobu version, as it's said — is a command. It means “you must mistake-proof.” It's forceful.
Mark Graban: So how does that get misinterpreted?
Martin Hinckley: A lot of times we take a casual view of mistake-proofing. Those who understand the problem realize it requires action, and it's expected — not optional.
Results: Mistake-Proofing vs. Six Sigma
Mark Graban: Can you give some examples of the benefits of poka-yoke and mistake-proofing compared to inspection or other quality control methods?
Martin Hinckley: We worked with Motorola after they had been doing Six Sigma for six years, and collected data from their companies — they were very gracious in supporting that work. After six years of Six Sigma, they were getting defect rates in the range of 1,000 parts per million. That's a lot higher than the three parts per million that Six Sigma projects. We found similar results at Texas Instruments and GE Aircraft Engines — both had been aggressively applying Six Sigma and were not achieving defect rates anywhere near the level they wanted.
Out of this study, Motorola was persuaded to try mistake-proofing on their Braintree product production line. They were able to get defect rates down to 100 parts per million — a tenfold improvement in quality.
In terms of controlling mistakes, there are three approaches. The first is simplification — most processes and products can have their complexity cut in half through fairly simple methods. Lean is one tool for simplifying a process, and you can get similar gains from simplifying products, tools, and equipment. That eliminates the opportunity to make certain mistakes entirely.
The second is mistake-proofing, which controls errors in the process.
The third — and I don't think many people appreciate this — is that most variation problems actually come from adjustments. A technique developed in Japan called Single Minute Exchange of Dies converts adjustments to settings, which eliminates about 98% of setup errors and virtually all adjustment errors. When Six Sigma groups were studying Japanese companies and found less process variation, they assumed it was coming from statistical methods. The truth is it was coming from SMED, which is a much simpler approach — no data collection required.
Mistake-Proofing Beyond Physical Devices
Mark Graban: Can you share examples of mistake-proofing processes that are less dependent on physical devices or product design?
Martin Hinckley: Absolutely. I was in a hospital one time watching a nurse change a catheter. The supplies were on the other side of the building. She started the process and made an error three times during catheter insertion — each time she had to go get a new kit, and each trip took about five minutes.
The right solution is to organize the kit so you can only draw the right material in the right order to insert the catheter. There are ways to structure the process so errors are prevented, not just caught after the fact.
Another example: a problem with catheters is that they're not removed at the proper time. If left in too long, they can result in a hospital-acquired infection. There's no way to look at a catheter or the tubing and know how long it's been inserted. A simple timer that starts when the process begins would address that kind of problem. Many times there are very simple things that could be done to mistake-proof generalized processes.
Structure of “Make No Mistake”
Mark Graban: Could you tell our listeners about the structure of your book?
Martin Hinckley: The biggest challenge with mistake-proofing is that it's not the way we've been trained to approach problem-solving. Our background and history as a society is based on the variation paradigm. Mistake-proofing is a skill — like playing chess. You get better by doing it.
The problem is that examples are scattered across a wide range of locations and industries. It's really hard to find an example that's relevant to your specific problem. So what we've done is develop a classification scheme — a way of organizing examples so you can find relevant solutions you might not have thought of on your own.
For example, an omitted part is the same type of problem whether you're in a hospital operating room or an automotive plant. The techniques that prevent omitted parts are the same in both environments. But if you show automotive examples to hospital staff, they often have difficulty translating them to their environment. So we've reduced each type of problem to a set of principles, with examples organized in a catalog format.
The introduction guides people into the method. Beyond that, we believe the best approach is for each organization to build its own database of examples — things people can share. We're developing software that will help companies create their own proprietary database. It will track every quality problem in the organization in a simple one-page before-and-after format, so people can see what was done and how it relates to their work.
Mark Graban: I would imagine that in industries where organizations tend to cooperate — like healthcare — there might even be opportunities to create industry-wide databases that people could reference and share.
Martin Hinckley: Yes.
Mark Graban: Is the software web-based?
Martin Hinckley: The intent is to make it web-based within the company. We'll also offer a version on the web with a smaller set of non-proprietary examples, so people can contribute and find examples that might help them.
Mark Graban: And is that being developed by Assured Quality?
Martin Hinckley: Yes.
About Assured Quality
Mark Graban: Can you tell listeners a little about your firm and how to get in touch with you?
Martin Hinckley: Sure. I'm president of Assured Quality. Our website is assuredquality.com — no spaces or punctuation. I'm also affiliated with QMI's Visual Lean Institute, and people can reach me through that website as well. Assured Quality is where the software is being developed.
Mark Graban: Listeners to the podcast may recognize the Visual Lean Institute and Gwendolyn Galsworth, who has been on here talking about visual management and visuality — she'll be back again in a future podcast.
Martin, thank you for joining us today. We were just able to scratch the surface of mistake-proofing, but I would definitely recommend the book to anyone who wants to learn more. That's available through Productivity Press, correct?
Martin Hinckley: Yes, definitely through Productivity Press.
Mark Graban: It was a pleasure having you on the podcast today. Hopefully we can have you back if listeners have questions or other topics they want to explore.
Martin Hinckley: Thank you so much, Mark. It's been my pleasure.
Is pokayoke all about preventing errors from happening in the first place or does it also include techniques to prevent errors from causing damages, such as rubberizing a wrench so if it slips it wont scratch or dent something, or a pan to catch accidental spills etc
I would classify rubberizing a wrench or spill pans as a part of the 6th S in the “5S Family”: Safety
Dave i though safety was protecting humans, that what most people think of, do lean people generalize it to also protecting parts or places from accidents too?
to Mike — I asked Dr. Hinckley to respond.. but my take is your examples would be a type of error proofing, something that lessens the impact of an error. Preventing the error from happening altogether would be ideal, but this isn’t always possible.
To Dave’s point… I wouldn’t call this “6S” since it’s not necessarily a safety example (a rubberized wrench would still hurt if it landed on your head) and I personally dislike taking “safety” onto 5S… just because it starts with S.
Safety is important, it’s critical. It should be a priority in everything, not just cutely tacked onto 5S, I think.
I got the safety piece from prevention of spills, people slipping, = safety.
Perhaps a more appropriate S in this case would be “shine”.
I find myself disagreeing that rubberizing and pans fit into the mistake proofing realm. In my mind, mistake proofing removes variance in a process so that quality improves.
Preparing for a mistake (rubber and pans in this example) is not poke-yoke (IMHO) because it is dealing with a mistake, a variance that has already occurred. Therefore, the action of dealing with the problem must be put into another lean philosophy bucket such as cleanliness or safety.
A better example of poke-yoke (IMHO) in these examples would be to make it so that the wrench won’t fall, and making it so that the spills won’t occur.
Fun stuff, thanks for reading my thoughts. =)
Dave – I agree with you that a spill pan is not the BEST error proofing. But sometimes you have to be practical. I’d define error proofing as anything that prevents a problem from occurring (physical defect or a process defect).
In medicine — if they’re trying to put a breathing tube down a patient’s airpipe, sometimes they’ll go down the foodpipe (pardon my non-technical terms). You can’t prevent this, at least with current technology. So they made the error easily detectable with a squeeze bulb. If the bulb doesn’t whistle, they know they’re in the foodpipe and they can try again.
I’d call that error proofing — it makes the error easily detectable so it can be reversed.
hmmm, I’d compare the whistle to an Andon light instead of a poke-yoke fixture.
The whistle is an indication that something went wrong and action is required. Poke-Yoke makes sure nothing goes wrong. I have a hard time deviating from the strictest definition when the subject of removing variances comes up.
You can call me a lean-fascist now, or “Lord Kanban”, as I am popularly known. =)
That’s the neat thing about lean to me, the answer is probably somewhere in the middle. We are only limited by imagination, and are by no means done growing.
Poka-yoke, or mistake-proofing must achieve at least one of four functions to be effective:
1. The product or process must be simplified to eliminate the opportunity for an error,
2. The error must be controlled or prevented,
3. The process must be shutdown, or
4. There must be a warning that the process is not adequately controlled.
Rubberizing a wrench so that it won’t scratch or dent an object is a form of mistake-proofing since it controls or prevents scratches or damage. However, there may be better control methods. To Illustrate, in some applications it may be possible to control the motion of the wrench, or it may be necessary to put a protective barrier between the wrench and the product. The latter solution can be very important if dropping the tool can damage the product or cause other safety problems. There is one case where a dropped tool caused the detonation of a rocket. Tools are now strapped to workers so that the drop distance is limited in such applications, but the attachment of the tool to the worker may not be mistake-proofed. In the electronic industry, a tool that slips off the fastener may destroy a circuit board, requiring a more robust barrier than coating the tool.
The key point is that the mistake-proofing solution generally has to be adapted to the specific task, and there is not a single solution that provides a complete description of the form of mistake-proofing required. Note, however, that each of these solutions (protective coatings, attachment of tools to the worker, and physical barriers between products and tools) all are based on the same principle – a barrier is placed between the tool and the product for protection.
With respect to spills, a better solution than a pan to catch spills (such as a pan under a washing machine) is determining the cause of spills and preventing them. Most of us are not happy if our car leaks oil, and putting a pan down to catch oil spills is not the ideal. Thus, while the pan may be considered a form of mistake-proofing it would be a very weak solution. We need to raise the vision with respect to what can be accomplished with mistake-proofing.
Martin Hinckley
Awesome examples Dr Hinkley.
Right now im a painting contractor so im a little crazy about spills… I sometimes run paper from the front door to the room im working in. Also I put my open paint cans and lids on top of something or of to the side more out of the way and not in the middle of a room or middle of a walking path(less-common sense?)
I do have more pokeyoke ish examples. I usually remove shutters to paint houses. Doing that on an extension ladder can be a little tricky to do and its easy to drop screws now and then, often they fall into plants/bushes and cant be retrieved. So I started using a magnetic wristband to collect/hold them. Not only have i not dropped a single screw nail or bit since, but the process is now faster and easier.
Mike, I think you examples are awesome too! Great Job. I apologize for not responding sooner. I have been on travel.
There are three levels of Poka yoke depending upon what it allows & prevents.
Level 3: Detects a defect after it has been made before it reaches the next operation.
Level 2: Detects an error in the process of it occurring, before it result in a defect.
Level 1: Eliminate error at the source, before it occurs.
Although Level 1 is best & in that hierarchy, it is not always possible & practical to go for Level 1 so depending on situation you can choose other levels also.
From this point of view I think the Rubberizing a wrench may be Level 3 but can be better solutions (as Dr. Martin has mentioned) at Level 1 & if not possible at least Level 2