Both stop defects. The difference is when — poka-yoke stops a defect from ever existing, jidoka stops one that already happened from going anywhere.
It's common to hear poka-yoke and jidoka lumped together as "the two Toyota tools for catching errors." That flattens a real distinction that matters for how you design a process. Poka-yoke ("mistake-proofing") is a design-level intervention: a part, fixture, or connector shaped so the wrong action is physically impossible to take in the first place. Jidoka ("automation with a human touch," sometimes translated "autonomation") is a detection-and-response principle: a machine or worker watches for an abnormality, and when one shows up, the process stops itself — immediately, before the defect can travel to the next station. One makes the error impossible. The other makes the error impossible to ignore or pass on. They usually work together, but they are not the same lever.
A poka-yoke device works even if nobody is paying attention, even if the operator is tired, rushed, or new — because the error was never an option the geometry allowed. That's its strength and its limit: it only works for failure modes you can design out physically. Jidoka covers the much larger space of abnormalities you can't design away — a tool that wears out mid-shift, a raw material that's slightly out of spec, a step a fatigued operator skips. Jidoka doesn't prevent that from happening once; it guarantees the process never stays silent about it. The line stops, the abnormality gets attention now, and — critically — it stops beforethe bad part reaches the next station, the next process, or the customer. In practice, mature lean systems use jidoka's stop-the-line discipline to build in poka-yoke devices over time: every abnormality jidoka catches is a candidate for a future poka-yoke that removes the failure mode for good.
They sit at different points on the timeline, and that difference changes what each one can promise. Poka-yoke is proactive and design-level: it makes the defect impossible to create or pass on in the first place — a connector that only fits one way, a fixture pin that only seats a correctly oriented part. Nothing is ever "caught," because nothing wrong ever happens. Jidoka is reactive-but-immediate and detection-based: something has already occurred — a torque out of range, a missing part, a dimension out of tolerance — and jidoka's entire contribution is refusing to let that abnormality travel any further, by stopping the process the instant it's detected. They are complementary rather than interchangeable, and in fact jidoka very often includesa poka-yoke device as its detection mechanism — the sensor or limit switch that tells the line "stop" is frequently a poka-yoke component doing double duty. The relationship runs one direction: every poka-yoke is a defect that jidoka never had to catch. Not every jidoka stop is a defect that could have been poka-yoked away.
Explains the real distinction between poka-yoke (mistake-proofing) and jidoka (automation with a human touch) — poka-yoke is a design-level device that makes a defect physically impossible to create or pass on, while jidoka is a detection-and-stop principle that catches an abnormality immediately after it occurs and halts the process before the defect propagates — using a mistake-proof fixture and an andon-triggered auto-stop as concrete examples.
Both terms come out of the Toyota Production System and both get summarized in training materials as "catches errors before they reach the customer," which makes them sound interchangeable. The summary isn't wrong, but it collapses a timing distinction that matters operationally. Poka-yoke intervenes before the fact — the wrong action is never physically available, so there is nothing to "catch." Jidoka intervenes immediately after the fact — an abnormality has already occurred, and jidoka's job is stopping it from traveling any further, not preventing its initial occurrence. Mixing them up leads teams to try to "jidoka" a problem that should have been poka-yoked (adding an inspection step for a defect that a simple fixture redesign could make impossible), or to assume a poka-yoke device by itself constitutes a jidoka stop-the-line system when it hasn't been wired into any stop/andon logic at all.
Poka-yoke devices fall into two broad types: contact methods, which use the physical shape, dimension, or other physical attributes of a part to detect or prevent an error (an asymmetric connector, a locating pin that only fits one hole pattern, a part that's too large to fit in the wrong fixture), and fixed-value/motion-step methods, which ensure a fixed number of movements or parts are used (a parts tray with exactly the right number of screw wells, a counter that won't release a pallet until the correct count is reached). The common thread is that no inspection, sensor reading, or human judgment call happens at the moment of use — the constraint is built into the physical or procedural setup itself, so the failure mode simply isn't reachable.
Jidoka is usually described as a four-step response: (1) detect the abnormality — a sensor, limit switch, or a trained worker notices something is wrong; (2) stop — the machine or line halts automatically, or a worker pulls an andon cord to trigger the stop; (3) fix or correct the immediate condition — clear the jam, replace the part, address what's in front of you right now; (4) investigate and address the root cause — often with a tool like 5-Why analysis — so the same abnormality doesn't recur. The first two steps are what most people picture when they hear "jidoka," but the last two are what turn a line stoppage into a real improvement instead of just repeated downtime. A jidoka stop that never gets root-caused just becomes an expensive way of catching the same defect over and over.
Yes, very commonly. Jidoka needs a detection mechanism as its first step, and a poka-yoke sensor or limit switch — one built to detect a specific physical condition, like a part not seated or a fastener not present — is often exactly what triggers the automatic stop. In that setup the poka-yoke device is doing the "detect" half of jidoka's detect-and-stop sequence.
Not necessarily automatic in the sense of no human involvement — jidoka literally translates closer to "automation with a human touch." Some jidoka implementations are fully automatic (a sensor trips a machine stop), while others rely on a worker recognizing an abnormality and pulling an andon cord themselves. Either way, the defining feature is that the process stops the moment something is wrong, rather than continuing to run and letting a supervisor or downstream inspection catch it later.
No. The principle — make the wrong action physically or procedurally impossible — shows up anywhere a process has a discrete failure mode with a clear physical or structural fix: a gas cap that won't let you drive off with the pump nozzle still attached, a software form that won't submit with a required field blank, a medical connector standard (like ISO 80369) designed so an IV line and a feeding tube physically can't be cross-connected.
In practice teams often build jidoka-style detection first, because it's usually faster to add a sensor and a stop condition than to redesign a part or fixture. But the more durable fix is almost always poka-yoke, because it removes the failure mode entirely rather than catching it every cycle. A common improvement pattern is: use jidoka to stop the line and study the abnormality, then design a poka-yoke device so that specific abnormality can't recur — turning a recurring jidoka stop into something that never needs to be caught again.
No. Jidoka triggers on an abnormality — a condition outside the expected range — not exclusively on a confirmed defect. A torque reading slightly under threshold, a cycle time that ran long, or a part that failed to seat correctly can all trigger a stop before any bad part is actually completed. That's part of the point: jidoka is often catching the precursor to a defect, not just the defect itself, which is what makes "immediately after" still fast enough to prevent the defect from ever fully forming or shipping.
Try our Industrial & Systems Engineering Studio
More calculators, simulators, and guides for this discipline.