A3 Problem Solving
One page. Six weeks. Watch the thinking happen.
An A3 is one page. That’s not a metaphor — Toyota engineers write the whole thing, problem to countermeasure, on a single A3-size sheet, and the name comes straight from the paper: 297 by 420 millimeters, the same size a mechanical drawing ships on. Six boxes live on it — background, current condition, analysis, target, countermeasures, follow-up — and you’ll watch all six fill in, over six weeks, on the sheet ahead.
One page is the whole design decision. A report can run twenty pages because nobody has to choose what matters — everything fits. An A3 doesn’t get that luxury. Every sentence competes for space with every other sentence, so the sheet forces you to decide what’s actually load-bearing and cut the rest. That constraint isn’t a formatting rule. It’s the method.
Which is also why an A3 is not a form. A form gets filled in top to bottom, once, then filed. An A3 gets read, questioned, erased, and rewritten — five or six times before anyone calls it finished. If nothing on your sheet ever gets scratched out, you probably wrote a report and drew six boxes around it.
The sheet has a geography, and it matters more than its contents. Find the fold down the middle: the left side grasps the situation — background, current condition, analysis. The right side acts on it — target, countermeasures, follow-up. That line is a discipline, not a decoration. Nothing gets written on the right side until the left side survives being questioned — hard, out loud — by someone whose whole job is asking “how do you know?” until you can answer, or you can’t. Watch for the veil over the right side below. It lifts when it’s earned, not when you’re ready to move on.
Look again and PDCA is underneath the boxes: Plan is the whole left side — background, current condition, analysis — plus the target on the right. Do is countermeasures. Check and Act are the follow-up. Now watch the sheet fill in.
The first draft is always confident. That's the problem.
A blank sheet, a rushed reflex, and a first draft built from what everyone already believes. Watch it get written — and then watch it get erased.
Line 3 housing defect rate — owner: M. Chen
The sheet starts blank. So does she.
Defects on Line 3 doubled — 4% to 8% — and the plant manager's reflex is already circulating: schedule refresher training. Instead, the problem lands with Maya Chen, a second-year engineer. Not because of her title; she doesn't have one worth mentioning. Because someone has to own the problem, and ownership — not rank — is what an A3 runs on.
She writes what everyone knows.
The first draft takes an hour. The problem statement proposes training. The current condition quotes what supervisors say. It's neat, plausible, and contains not one number anyone measured. Most A3s die here — approved.
Diane doesn't correct it. She questions it.
Three red-pencil questions land in the margin. How do you know? Where would I see this? Is this a problem — or a solution wearing a disguise? A coach who writes on your A3 owns it. A coach who questions it makes YOU own it.
The best thing on this A3 is what leaves it.
Maya can't answer the questions. That's the answer. The opinions come off the sheet — every unproven line of them. Erasing feels like failure; it's the first real progress. The sheet is honest now: it says almost nothing, which is exactly what she knows.
Which one is a problem statement?
Maya wrote all three before Diane read any of them. Tap each to hear the coach’s reaction.
The facts aren't at anyone's desk.
Five facts land on the sheet over three weeks — not researched, but walked to. Quality, maintenance, receiving, the press itself: each fact has a name attached, because someone had to go and ask.
Line 3 housing defect rate — owner: M. Chen
The training theory dies in one chart.
Marcus Santos in quality pulls the defect data by shift: A 7.9%, B 8.2%, C 7.8% — nearly identical. If undertrained new hires were the cause, one shift would carry the blame. None does. The theory that seemed obvious in the plant manager's office doesn't survive five minutes with the actual numbers.
The defect has an address.
Not every part is failing — just one. Marcus narrows it further: defects cluster in the cast housing, dimensional, out of tolerance right at the press. A vague "defects are up" just became a specific part, a specific station, a specific kind of failure. That specificity is what lets an A3 stop guessing and start pointing somewhere real.
It has a birthday, too.
Twelve weeks of quality data, laid end to end, show something training theory can't explain: the rate doesn't creep — it steps. Flat at 4% through early March, then a jump to 8% the week of the 12th, holding ever since. A defect with a birthday points to a change, not a skill gap. Something happened that week.
Maintenance has been waiting for this question.
Dave Cooper in maintenance didn't need to be asked twice. Press calibration ran weekly until February, when a schedule revision stretched it to monthly. "We flagged it in February," he says. "Nobody asked." The fact was sitting in a maintenance log the whole time — waiting for someone to walk over and ask the person who'd know.
Receiving finishes the picture.
Jim Walsh in receiving supplies the last piece: the casting supplier changed on 3/10 — two days before the timeline's own turn. New lots pass inspection, technically in-spec, but sit at the wide end of tolerance. Brian Reed, running the press, already knew it in his hands: "They seat differently. We adjust by hand when there's time." Go see, and the plant tells you everything.
Sort the search onto the fishbone.
Six facts came back from three weeks of asking around. Tap a fact, then tap the bone — Man, Machine, Material, Method, Measures, Environment — you think it belongs on.
Tap a fact, then tap a bone to place it.
0 of 6 on the bone we’d pick. The exact bone is a judgment call; a complete search before judging any single cause is the point.
- Casting supplier changed 3/10; lots at wide end of toleranceMaterial — variable input is half of the compounding pair.
- Calibration cadence cut weekly → monthly in FebruaryA method/standard decision — the schedule change, not the machine, is the cause.
- Press drifts out of true between calibrationsMachine — the drift is real, but ask why it’s allowed to persist.
- Defect data was only reviewed monthly, so the step-change hid for weeksMeasures — slow feedback let a fast problem look gradual.
- Operators compensate by hand, each differentlyPeople adapting to a broken system — a symptom to relieve, never a cause to blame.
- Order volume was flat all quarterEnvironment — checked and ruled out. Ruling things out is analysis too.
The right side stays locked until it's earned.
Analysis, a near-miss with blame, and then — finally — a target, a tested countermeasure, and a sheet that makes its way around the plant before anyone approves it.
Line 3 housing defect rate — owner: M. Chen
Two causes, one problem.
Six categories, run straight down the fishbone, narrow to two: MATERIAL — the wider casting tolerances since the supplier change — and MACHINE/METHOD — a press no longer trued weekly. Neither one, alone, explains an 8% defect rate; the plant has absorbed worse. Together, compounding, they do. Root cause isn't a single culprit. It's two "acceptable" changes meeting at the same station.
A name almost goes on the sheet.
The draft almost says it: operators compensating inconsistently by hand. A name is one keystroke away. Diane's pencil stops it cold — no names, no "operator error." If someone is struggling, the system failed them; what forces the compensation? The line comes off the sheet, struck, and gets rewritten as a condition, not a person.
The veil lifts.
For five weeks the right side was locked. Nobody argued about solutions, because there was nowhere to write them. Now the target lands first, not a fix: back to ≤4% by May 30, measured weekly at final inspection, no added inspection labor. A target with a number and a date is the only thing allowed to earn the space a countermeasure will eventually fill.
One countermeasure dies, one gets tested.
The easy fix goes up first — switch back to the old supplier — and dies fast. Diane: it treats one cause, prevents nothing, adds cost; what does the pairing tell you? What survives is a test, not an opinion: Dave and Maya restore weekly calibration on one press for one week. Defects on that press alone fall from 8.1% to 4.6%. Now there's a countermeasure with evidence behind it.
The sheet goes for a walk.
Three countermeasures go up — calibration cadence to maintenance planning, tightened incoming spec to purchasing, standardized press setup written with the operator who runs it — and the sheet leaves Maya's desk. Quality adds a sticky: measure weekly, or we'll argue about the number. By the time this reaches approval, it takes ten minutes; it's the last people to see the A3, not the first.
v5 — the residue of thinking.
Thirty days out, the check: 3.8%, under standard, with a 60-day check already on the calendar. The calibration cadence lives in the TPM standard now; the incoming spec lives in the supplier agreement. Pull back and look at the whole sheet — six regions, five versions, more strikethroughs than clean lines. That residue is the thinking. The paper was never the point.
How to write an A3 problem statement
A problem statement earns its two lines by containing exactly four things: the gap — what’s happening versus what should be; the standard it’s measured against; the magnitude — how big, in numbers; and the when/where — since when, on what line, in what part. Maya’s rewritten line has all four: housing defect rate on Line 3, 4% to 8%, beginning the week of March 12, standard ≤4%. Read it again and notice what’s absent — no cause, no fix, no name. That absence is the point.
The three drafts Maya considered map onto the three ways a problem statement usually fails. “Operators need better training and closer supervision” isn’t a problem statement — it’s two solutions wearing a disguise, and it tells you nothing about what’s actually wrong. “Defects are unacceptable and must be reduced as soon as possible” fails differently: no gap, no standard, no timing — a sentence that fits every problem describes none of them. And the version everyone reaches for first, built from what supervisors already believe, fails the way most confident first drafts do: plausible, unmeasured, wrong.
The rule that catches all three: if a sentence contains a cause, a fix, or a name, it doesn’t belong in the problem statement — keep erasing until it’s gone. “Operators need training” is a fix. “Because new hires are rushing” is a cause. Either one, and you’ve stopped investigating before you started — the rest of the A3 becomes a formality that confirms what you already decided in the first paragraph.
Here’s the part that’s easy to miss: the problem statement isn’t written first, even though it sits at the top of the sheet. It’s written after the current condition — after you’ve gone to see the data, the shift breakdown, the timeline — not before. Write it too early and you’re describing your assumption. Write it after, and you’re describing what you actually found.
The questions an A3 coach asks
Diane never once wrote on Maya’s A3. Every red-pencil mark in this guide is a question, not a correction — and that’s not a stylistic choice, it’s the whole coaching method. A coach who edits the sheet ends up owning the thinking; a coach who only asks questions sends the owner back to reality, again and again, until the sheet can survive them on its own. Keep these in your pocket:
- “What problem are you trying to solve — in one sentence, no causes, no fixes?”
- “How do you know? Where would I see it?”
- “What is the gap, against what standard, since when?”
- “Who have you talked to that touches this every day?”
- “Is this a problem, or a solution wearing a disguise?”
- “What would have to be true for that cause to produce THIS data?”
- “What did you rule out, and how?”
- “What’s the smallest test that could prove you wrong?”
- “Who has to agree before this works — and have they held the pen yet?”
- “When will you check, and what number will tell you it worked?”
None of them are gotchas. They’re the same handful of moves, aimed at whatever’s currently on the sheet: press on the evidence, press on the scope, press on the causal story, press on what happens next. Ask them of your own A3 in that order and you’ll find, fast, which box isn’t ready yet.
One of those questions is really a rule with its own name: the anti-blame rule. Watch what Diane does the moment a name almost lands on the sheet — “operators compensating inconsistently by hand” — she doesn’t let it stand. No names, no “operator error.” If someone is struggling inside a system, the system failed them first, and the question that follows isn’t who — it’s what forces the compensation. A coach who lets a name onto the sheet has stopped investigating a system and started assigning blame, and the A3 stops teaching anyone anything.
Notice, one more time, who owned this A3: not the plant manager, not Diane — Maya, a second-year engineer with no title worth mentioning. Ownership on an A3 was never about rank. It’s about who is willing to go and see, sit with the questions, and carry the sheet around until it says something true. Title tells you who signs it. It doesn’t tell you who wrote it.
By the time this A3 reached anyone’s desk for approval, purchasing, quality, and maintenance had already added their own sticky notes to it — which is the actual point of the walk in Act III. An A3 isn’t a report that gets approved; it’s a consensus object that gets built, department by department, until approval is ten minutes of paperwork on thinking everyone already agrees with.
And most A3s that die don’t die at countermeasures — they die in the follow-up box, the one nobody checks. That’s where a problem actually ends: not when a fix ships, but thirty days later, when someone looks at the number again.
Line 3, its people, and every number on this A3 — the defect rates, the timeline, the costs — are illustrative, invented for teaching. The method is real.
Further reading: J. Shook, Managing to Learn (Lean Enterprise Institute, 2008) · Lean Enterprise Institute, “A3 Report” lexicon.
Frequently asked
- What is A3 problem solving?
- A method from Toyota for working through a problem on a single A3-size sheet: background, current condition, analysis, target, countermeasures, follow-up. The paper is the visible part; the method is the thinking it forces — grasp the facts before proposing fixes, find causes in the system rather than in people, and build agreement by carrying the sheet to everyone it touches before any approval meeting.
- How do you write a good A3 problem statement?
- State the gap against a standard, with magnitude, location, and timing — and nothing else. "Defect rate rose from 4% to 8% beginning March 12; standard is 4%" is a problem statement. "Operators need more training" is a solution wearing a disguise, and it ends the investigation before it starts. If your statement contains a cause, a fix, or a name, it isn't done.
- What questions does an A3 coach ask?
- Questions that test whether you've actually seen the problem: How do you know? Where would I see it? What did you rule out? What's the smallest test that could prove you wrong? Who touches this every day, and what did they say? A good coach never corrects the sheet directly — the questions send the owner back to reality until the sheet answers them.
- How is an A3 different from PDCA?
- It isn't separate — an A3 is PDCA made visible on one page. The left side (background, current condition, analysis, target) is Plan; countermeasures are Do; follow-up is Check and Act. If your A3 ends at countermeasures with no follow-up box, you have half a PDCA cycle and no learning.
- How long should an A3 take?
- Days to weeks, not hours — most of it away from the desk. The writing is fast; the going-to-see, measuring, testing, and socializing are the work. An A3 finished in an afternoon almost always contains what everyone already believed, which is exactly what it exists to protect you from.
Related
Founder of Kaizumi, an AI-powered Lean training platform. More about Matthew →
Updated July 2026 · Line 3, its people, and all figures are illustrative, created for teaching — the same case you can investigate yourself in the A3 Investigator. The method — one page, facts first, causes in the system, consensus before approval — follows Shook, Managing to Learn (Lean Enterprise Institute, 2008).