Watch the recording of the webinar:
Listen to the recording via our podcast:
View the Slides:
Alternative download (PDF)
Most organizations respond to recurring errors with the same set of moves. Tell people to be more careful. Add an inspection step. Update the training. Send a reminder. Post a sign. Then the same error happens again, often with a different person making it.
Poka-yoke — mistake proofing — is the alternative. Instead of asking people to be more vigilant, you change the design of the work so the right action is the easiest action and the wrong action becomes difficult or impossible.
Dr. John Grout has spent his career on this topic. He holds the David C. Garrett Jr. Professor chair at Berry College and won the Shingo Prize for his paper "The Human Side of Mistake-Proofing," co-authored with Douglas Stewart. He has consulted with organizations across industries on designing errors out of their processes and is one of the most-cited researchers in the field. The session is delivered in PechaKucha format — 20 slides at exactly 20 seconds each, twice — which means the format itself mistake-proofs against running long. The pacing is fast. The examples are dense. The principles stick.
What follows is the substance of the session, organized so you can use it whether you watched it or are landing here from search.
The term comes from Shigeo Shingo, but the concept predates the term by centuries. The poison bottle from the 1800s John shows is the proof: color-coded, spiked, designed so a pharmacist working in low light would notice it through both vision and touch. The principle is the same then as now — make the right action easy and the wrong action obvious or impossible.
The English translation most commonly used is "mistake proofing." A few translations to be careful of: poka-yoke is sometimes mistranslated as "fool-proofing," which Shingo himself rejected. The principle isn't that workers are fools. It's that humans — all humans, including the most skilled — make predictable errors under predictable conditions, and well-designed processes accommodate that reality rather than fighting it.
The shift in framing is what makes poka-yoke powerful. The question is not "why did this person make this mistake?" The question is "what about the design of this process allowed the mistake to be possible?"
Mistake proofing operates at three levels.
Detection. The error is caught at the moment it occurs. John's example from the session: orange dots painted across the heads of bolts on a flange. If a bolt is under-torqued, the dot's position shifts and the misalignment is visible at a glance. The detection only works if the response works too — catching the error means nothing if no one notices or acts on the signal.
Prevention. The error cannot happen at all. John's most memorable example: the SawStop table saw with a system that senses human flesh and stops the blade in milliseconds. The video clip he shows uses a hot dog — the blade snaps out of the way fast enough that the hot dog comes out with a small nick instead of being severed. The cost: a $70 single-use SawStop module and a $75 saw blade. Top-of-the-line saws with this technology cost $300 to $700 more than a typical competitor. John's framing: you have to decide how much your finger is worth.
Fail-safing. When prevention and detection both fail, the consequences are minimized. The breakaway gas pump hose John shows is designed to separate cleanly when a driver pulls away with the nozzle still attached. Elevator brakes — Elisha Otis's actual invention, John points out, wasn't the elevator but the elevator brake — are the same principle. The cable can fail. The car will not fall. Otis's 1853 invention is what made modern skylines possible.
The order matters. Most organizations default to detection (inspection, checklists, audits) when prevention would be cheaper, more effective, and less burdensome. Designing the failure out of the process is almost always better than designing a system to catch the failure after it occurs.
A foundational concept John draws from cognitive psychologist Donald Norman: put knowledge in the world rather than knowledge in the head.
The example: a Toyota engine where the dipsticks are labeled "T" for transmission and "O" for oil. The labels live in the world, on the parts themselves, where the user encounters them at the moment of decision. Knowledge in the head — remembering which dipstick is which — is unreliable. Knowledge in the world — looking at the letter — is durable.
The principle generalizes. Knowledge in the world can communicate when something is worn out, what type of object it is, what temperature something is, how to get somewhere. The toilet seat with red lights when the seat is up and green lights when it's down keeps people from falling in at night. The principle is small and the application is universal.
Norman's The Design of Everyday Things is the canonical reference. John recommends it directly. If poka-yoke as a concept clicks for you, that book is where to go next.
This is the most operationally useful single section of the session for a CI leader trying to apply poka-yoke effectively.
John draws on the work of cognitive psychologist James Reason, who classified human error into three types based on the cognitive level at which the error occurs. The classification matters because different error types respond to different countermeasures.
Skill-based errors happen during routine, automatic actions. Reason's example John walks through: have you ever driven through a traffic light and not remembered whether it was green? That's skill-based action. No deliberation, no decision-making, automatic execution. Errors at this level are slips — you intended the right thing, your execution failed.
Rule-based errors happen when you recognize the situation and apply a rule. The example: you come to a blinking red light. You recognize the unusual circumstance, retrieve the rule from your driver's training (treat it like a stop sign), and apply it. Rule-based errors happen when the wrong rule gets retrieved or the rule doesn't actually fit the situation.
Knowledge-based errors happen in genuinely novel situations where no rule applies and you have to reason from first principles. The example: the light turns red as you approach, you wait for several minutes, nothing happens, no rule covers this exact situation. You have to think — about your day, about an alternate route, about whether running the light is safe and whether you'll get a ticket. There's no predetermined correct answer.
The key implication for poka-yoke, in John's words: if the correct answer is not known in advance, you probably cannot use mistake proofing.
Mistake proofing works very well for skill-based errors and rule-based errors, where the correct action is known and definable. You can design environments where the wrong action is difficult or impossible. For knowledge-based errors — diagnoses, novel engineering problems, situations requiring genuine judgment — mistake proofing has limited applicability. Better information, better collaboration, and better decision-support tools are more appropriate.
Most recurring errors in operational settings are skill-based or rule-based, which is why poka-yoke produces such large returns. The errors are predictable. The design changes that prevent them are usually low-cost. The improvements compound.
A heuristic John uses to clarify what counts as mistake proofing.
The general rule: if you can't take a picture of it, it's not mistake proofing. Tangible design changes — physical fixtures, color coding, shape constraints, sensor-driven mechanisms, software interlocks — are the substance. Policies, training, reminders, and "be more careful" exhortations are not mistake proofing, no matter how well-intentioned.
The exception John names: high-tech equipment or software where the mistake-proofing logic exists in code rather than as a physical object. His example: a phone that stops audio output anytime the Bluetooth speaker is turned off, so you don't miss any of your song or audiobook. You can't photograph the logic, but the logic is real and the protection works.
The rule is useful as a filter when teams are evaluating what they're calling poka-yoke. If the proposed countermeasure is a sign, a meeting, a reminder, or a memo, it isn't poka-yoke yet. The next question is what physical, visible, design-level change would address the same error.
A reframe John uses that changes how teams approach problems where prevention is hard.
Sometimes you can't prevent the error directly. What you can do is replace the failure mode with a less harmful one. Otis's elevator brake is the canonical example: getting stuck between floors is a failure, but it's a far better failure than falling.
John gives the principle a name and a tongue-in-cheek acronym: Benign Failure Design — BFD. The framing matters because it expands the design space. Instead of asking "how do we prevent this error," you ask "if this error is going to happen anyway, what failure mode would cause the least harm?"
The star-shaped wheel that only turns when its container is right-side-up is an example. The wheel jams when the container is upside-down. The machine fails. But the machine failing is much better than the wrong product flowing downstream.
The copy machine bump that prevents the door from closing when the blue lever is open is another. The bump fails to allow the door to close. The user is forced to close the lever first. The failure protects the user from the worse failure of running the machine in an unsafe configuration.
A practical methodology section. Brainstorming alone produces ideas. Try-storming — brainstorming combined with rapid physical prototyping — produces working solutions much faster.
John walks through Boeing's "moonshine shops" as an example. Small groups of talented tinkerers with low supervision and high autonomy, given budget and physical space to mock things up with cardboard, tape, scrap parts. They prototype quickly, discard what doesn't work, refine what does, and finish with polished improvements at extremely low cost.
The specific example he shows: a tool the moonshine shop redesigned where the weight dropped from 22 pounds to 2.5 pounds and the installation time dropped from 30 seconds to 2 seconds. Total out-of-pocket cost: about $2,000 plus a little labor.
The other moonshine principle worth lifting: get out of Kansas. Leave the work area. Look at the world. The seat-loading elevator for 757 airliners came from someone seeing a hay elevator on a farm. Cut installation time for a full set of seats from 12 hours to 2 hours and freed up the overhead crane for other work.
The two design principles together — try-storm fast and look outside your industry — are John's recommended starting point for any team beginning serious poka-yoke work.
A specific rule John gives during the session, drawing from his consulting practice.
Before settling on a mistake-proofing solution, generate at least seven candidate ideas. Seven isn't magic. It's a lower bound that forces you past the obvious first answer. The first idea is rarely the best one, and brainstorming the second, third, fourth, fifth, sixth, and seventh stretches the imagination enough to land somewhere genuinely good.
This pairs with the try-storming principle. Generate seven ideas. Mock up the most promising three quickly. Test them. Refine. The team that commits to the first idea usually ends up with a solution that works but isn't optimal. The team that generates seven and prototypes finds the design that prevents the error and speeds up the work and costs less.
A point John returns to throughout the session. The best mistake-proofing devices don't just prevent errors — they make work faster, easier, and smoother.
His framing: if a mistake-proofing device slows people down, they will find workarounds. People will route around friction under pressure. Mistake proofing that makes the work harder is mistake proofing that gets bypassed.
The Lockheed Martin "hog trough" he describes is a model example. C-130 wiring harnesses used to be coiled, carried to the aircraft, uncoiled, and connected — a process where the connectors caught on each other and produced loose connections. The hog trough is a custom carrier that holds the wiring harness with all connectors poking out where they should be. Workers carry the trough into the plane, screw it in place, plug everything in. The error mode disappears. The process is also dramatically faster — less troubleshooting, no unrolling, fewer reconnections.
The lesson: the best poka-yoke designs eliminate the error and reduce cycle time at the same time. When that happens, adoption is automatic. People use the tool because it makes their work easier, not because they were told to.
A point that gets missed in most introductory poka-yoke material and that John raises explicitly during the Q&A.
A sensor that's supposed to detect defects but hasn't been calibrated in six months provides false confidence, not quality. A pick-to-light bin system with a broken light is worse than no system, because operators trust the lights.
John gives a specific practice from a Honda supplier he works with. Next to each machine they keep parts painted red — what they call "poka-yoke parts." These are deliberately defective parts, designed to be just barely wrong. Once a shift, operators run them through the mistake-proofing system to verify the sensors still detect the defects. If the sensors miss the bad parts, the system needs maintenance.
The principle generalizes. Any mistake-proofing device that depends on sensing, reading, scanning, or measuring needs verification that the sensing still works. The verification needs to be part of standard work, not something that happens when someone remembers.
A list John walks through from Shingo's original cataloging, with practical context for each.
Guide pins. Pins that physically locate parts so they can only be assembled in the correct orientation. Common in machining and assembly.
Blinking lights and alarms. Standard equipment in most factories. Most useful when paired with an immediate response — a light that blinks but no one acts on isn't really mistake proofing.
Limit switches. Detect whether a part is present or absent. Simple, robust, widely used.
Proximity switches. Detect whether something is within a defined zone. Used for safety interlocks, position verification, and presence checks.
Counters. Verify the right number of operations have been performed. The Hella Windows hardware kit John describes uses this principle — a backlit board with holes for each piece of hardware. Any unfilled hole lights up, signaling missing components.
Checklists. John has mixed feelings about these, expressed directly. Most people complete the work first and then check the checklist, which loses most of the value. The most effective checklists are embedded in the actual process — like the Hella backlit hardware board — rather than completed retrospectively. For more on this distinction, John recommends Atul Gawande's The Checklist Manifesto.
A consolidated list of the examples John uses or mentions across the session, useful as a reference.
The 3.5-inch floppy disk has eight possible orientations and only one fits — a textbook prevention design. (John notes the icon for "save" in most software is still a floppy disk, which dates the example but keeps the principle live.)
Modern gas caps are tethered to the filler neck so they don't get left at gas stations. Same principle for electric vehicle charging caps.
Lawnmowers require pulling back a second handle while starting them, accounting for both hands and making it harder to cut off your fingers.
Glidden's ceiling paint goes on pink and dries white, so you can see exactly where you've already painted on a white ceiling.
Pick-to-light bin systems light up the correct bin and sound an alarm if the operator reaches into the wrong one. Off-the-shelf, networkable, configurable for custom pick lists.
Plastic geometric markers on truck lug nuts make it visually obvious if any nut has loosened — the pattern breaks before the failure does.
Highway rumble strips don't prevent the lane drift but detect it fast enough that the driver can correct before the failure becomes serious.
Round manhole covers can't fall through their hole, no matter how they're rotated. Rectangular ones can. The geometry mistake-proofs against worker injury below.
Cars now require the brake pedal to be depressed before the ignition will engage, eliminating a category of errors where drivers attempted to start in gear.
ATMs release the card before dispensing cash because data showed people walked away with the cash and forgot the card. Reordering eliminated the failure mode.
Surgical patients sign their surgical site while awake, so the surgeon has direct evidence of the correct location before incision.
The fly etched into the porcelain at JFK's urinals reduced spillage by 80% — though John shows a counter-example where the fly was placed too high to be visible during use, illustrating that first attempts at design aren't always best.
He says it directly. The homework: implement one mistake-proofing device.
Eliminate one error. Pick something small. Make the design change. Document what happened.
His framing: once you've done one, the rest are easy. The leverage is enormous because mistake proofing typically costs little and the savings can be profound. He references a Johnson & Johnson employee who eliminated a $14,000-per-year defect using Post-it notes.
The second assignment: do something based on the moonshine principle. Try-storm. Use whatever is on hand. Build something cheap and good. Be proud of it.
A few specific things the platform does that connect to systematic poka-yoke practice.
KaiNexus tracks improvement work across an organization, which means mistake-proofing solutions developed at one site become visible to every other site that could benefit. The most common gap in mistake-proofing programs isn't generating good ideas — it's spreading the ideas that work.
The platform supports the impact tracking that justifies continued investment. Errors prevented, defects avoided, time saved — the data is captured alongside the implementation.
It also supports the ongoing maintenance discipline mistake-proofing devices require. Audits, calibration checks, verification routines all live in the same system as the original implementation, which means they actually happen.
If your mistake-proofing efforts are producing local wins but not spreading, or if good designs are being implemented but not maintained, the gap is usually the infrastructure. That's the gap KaiNexus is built to close.
Dr. John Grout is the David C. Garrett Jr. Professor, former dean, and award-winning teacher in the Campbell School of Business at Berry College. He has researched lean supply chain management and mistake-proofing (poka-yoke) extensively and has published numerous articles on the topic. John was awarded the Shingo Prize for his paper "The Human Side of Mistake-Proofing," co-authored with Douglas Stewart. He has consulted with a wide variety of firms to mistake-proof their processes and maintains the resource site mistake-proofing.com, where his book Mistake-Proofing the Design of Health Care Processes — published by the Agency for Healthcare Research and Quality — is available as a free download.
What is poka-yoke?
Poka-yoke is the Japanese term for mistake proofing — designing processes so that errors are prevented, detected immediately, or made harmless when they occur. The term was coined by Shigeo Shingo as part of his work on the Toyota Production System, though the underlying concept predates the term by centuries. The principle is to make the right action easy and the wrong action difficult or impossible, rather than relying on vigilance, memory, or training.
What's the difference between prevention, detection, and fail-safing?
Prevention stops the error from occurring at all (a SawStop blade that stops on contact with flesh). Detection catches the error at the moment it occurs (orange dots that show whether bolts are properly torqued). Fail-safing minimizes the consequences when prevention and detection both fail (a breakaway gas hose, an elevator brake). Prevention is the strongest because it eliminates the failure mode entirely. Most organizations default to detection when prevention would be cheaper and more effective.
Why doesn't telling people to be more careful work?
Because vigilance, memory, and attention are fragile controls. They degrade under fatigue, pressure, distraction, and routine. James Reason's research shows humans make predictable errors under predictable conditions — and asking them to override that reality through willpower is asking the impossible. Mistake proofing works because it changes the environment so the right action is the easiest action, which doesn't depend on cognitive consistency that humans can't reliably maintain.
What types of errors does mistake proofing work for?
It works very well for skill-based errors (routine actions where execution slipped) and rule-based errors (where the wrong rule was applied to the situation). For these, the correct action is known and definable, and the environment can be designed so the wrong action is difficult or impossible. Mistake proofing works less well for knowledge-based errors (novel situations requiring genuine judgment), where better information, decision support, and expert collaboration are more appropriate.
What does effective mistake proofing actually look like?
Generally tangible design changes to the physical or digital environment. John's heuristic: if you can take a picture of it, it's probably mistake proofing. The best solutions are low-cost, low-tech, and highly effective — and they tend to speed up the process rather than slow it down. Examples include connectors that fit regardless of orientation, surgical checklists with verbal verification, pre-filled medication syringes, brake-pedal interlocks for car ignitions, and ATMs that release the card before dispensing cash.
Why should you generate seven ideas before committing to a solution?
Because the first idea is rarely the best. Seven isn't magic — it's a lower bound that forces you past the obvious first answer. Stretching the imagination through the second, third, fourth, and fifth ideas usually produces meaningfully better designs. Pair this with try-storming (brainstorming combined with rapid physical prototyping using cardboard, tape, and scrap parts) and the team will land on a design that prevents the error, speeds up the work, and costs less than the first idea would have.
Do mistake-proofing devices need maintenance?
Yes. A sensor that hasn't been calibrated in six months provides false confidence rather than quality. John describes a Honda supplier that keeps deliberately defective "poka-yoke parts" next to their machines and runs them through the mistake-proofing system once a shift to verify the sensors still detect the defects. The principle generalizes — any mistake-proofing device that depends on sensing, scanning, or measuring needs verification, and the verification should be part of standard work.
How do you mistake-proof in service industries?
Same principles, different applications. John's example from his consulting work: a company having problems with sales reps quoting impossible ship dates (December 25, July 4) to customers. The solution was a simple job aid — a calendar of valid ship dates posted around each rep's monitor. Healthcare has extensive mistake proofing through pre-filled syringes, surgical site marking, computerized medication dispensing, and pre-procedure verification protocols. In services, customer-facing mistake proofing matters too — the swing arm at the front of school buses that forces children far enough forward to be visible to the driver is a classic example.
.svg)
Copyright © 2026
Privacy Policy