A3 Problem Solving

Goal

A3 problem solving gathers on a single page, in a fixed order, the problem, its analysis, the proposed response and the plan that carries it. The deliverable is the filled-in sheet: choosing what to cut so that it fits is already analysis work.

The sheet makes departments with different angles on the same problem converge. It also carries a request for agreement: the owner goes out and wins each party's assent to the countermeasures before committing them. It is a mentoring medium too, the manager questioning his colleague's sheet instead of handing over the answer. Shook sees the technique's greatest benefit there: a management process that teaches through the scientific method.

Usage

When to use it

• Cross-functional or strategic problem at management level: the most frequent use at Toyota, according to David Verble of the Lean Enterprise Institute.
• Strategy deployment and planning of a major project: one sheet per level gives management and the teams a shared artifact.
• Departments disagree on what the problem is: the single page forces them to settle the current conditions together before discussing solutions.
• Decision to escalate to a committee: a page that lays out the reasoning can be challenged, where a long slide deck is endured.
• Manager willing to question rather than correct: mentoring through the sheet bears on an open question, inside his colleague's remit, whose answer is still to be found.
• Organisation that already documents its problems in long reports: the one-page limit shows what the volume was hiding.

When not to use it

• Incident to contain within the hour: apply the incident procedure or the eight disciplines (8D), whose third step isolates the problem from the customer before the search for causes.
• Data-driven improvement project on an instrumented process: follow DMAIC, whose five chained phases close on standard procedures.

Description

One sheet, two columns

The idea comes from Toyota: any question an organisation meets can and must fit on a single sheet, so that everyone concerned sees it from the same angle. The template published by Shook and David Verble organises that sheet into two columns. The left column carries the title, then sections I to IV, which establish what is happening: background, current conditions, goals, analysis. The right column opens with the owner and date field, then runs sections V to VII, which carry the response: proposed countermeasures, plan, followup. An arrow links each section to the next inside its own column, and the whole left column feeds section V. Whoever reads the page top to bottom then left to right walks the reasoning in the order it was built.

Shook writes that no single fixed template is the right one: the author decides what to bring forward according to the situation. Organisations adapt the layout, the headings and the relative weight of the sections. The sequence stays the same: a problem, its causes, a target, actions and the way the result will be judged.

The seven sections

Header fields come before the numbered sections. The title names the problem, the theme or the question being treated. The owner and date field says who carries the question and gives the date of the last revision, which already tells the reader that the sheet gets picked up again.

I. Background establishes the business context and the importance of the question. Shook puts the control question there: why are we talking about this?

II. Current conditions describes what is known today about the problem. The template asks for that state to be shown through charts, drawings or diagrams, then for the problem to be stated. It is the most expensive section, because it demands going out to observe the work and to put numbers on it, through value stream mapping where that is what it takes.

III. Goals identifies what is expected: which specific outcomes are required? A quantified, dated goal makes section VII verifiable.

IV. Analysis works back to the causes of the gap between the current situation and the wanted result. The template gives one instruction on tooling: take the simplest analysis tool that clearly shows the cause-and-effect relationship. Root cause analysis supplies that tool: the five whys when a single causal chain is plausible, the fishbone diagram when several families of factors are at play.

V. Proposed countermeasures lays out the proposal for reaching the target state and explains by which mechanism each countermeasure acts on the cause identified in section IV.

VI. Plan prescribes who does what and when, then names the performance or progress indicators. The template asks for a Gantt chart or its equivalent to be built into it.

VII. Followup anticipates the difficulties ahead, keeps PDCA going, captures and shares what has been learned. Verble names the outcomes of the checkpoint: standardise what held, continue the resolution if the gap remains, restart the sheet from scratch if the target was wrong.

The ten questions that guide the writing

Shook gives ten questions that the format and the goals of an A3 answer. They govern the writing, section by section, and are put as much to the author as to his reviewer.

1. What is the problem or the question?
2. Who owns it?
3. What are its root causes?
4. Which countermeasures could be considered?
5. How will you decide which ones to propose?
6. How will you get agreement from everyone concerned?
7. What is your implementation plan: who, what, when, where, how?
8. How will you know whether your countermeasures are working?
9. Which followup difficulties do you anticipate, and which problems may arise during implementation?
10. How will you capture and share what has been learned?

The mechanism is the dialogue

Shook presses one point that the sheet does not show: the reasoning it makes visible assumes a continuing exchange between the owner of the question and the other people in the organisation. He writes that the thinking which carries the participants through a PDCA cycle is what counts and not the form of the report. So the owner takes one version of section II to the people who do the work, a version of section IV to those who know the technique, a version of section V to those who will have to apply it. Each pass corrects the sheet, and the date field carries the last of those corrections.

Verble puts it the other way round: an A3 report, like an 8D report, describes problem solving once it has taken place; filling in the report solves nothing.

The mistakes that empty the exercise

Arriving with the answer. Shook documents this as the beginner's reflex. An author's first draft betrays his need to show that he has a solution: it races through sections II to IV to reach section V. The symptom is visible to the naked eye, the left half of the sheet empty and the right half full.

Writing current conditions without observation. Three lines of opinion in section II make the analysis unverifiable and the target arbitrary. The template asks for that state to be shown, which assumes it has been measured.

Proposing orphan countermeasures. An action that attaches to no cause in section IV reveals that the analysis was written after the fact, to dress up a decision already taken.

Keeping the paper and losing the constraint. A sheet printed in 8-point type, completed by four appendices, has kept the format and abandoned what makes the author work. The check: can the page be read at arm's length, standing up, by somebody who has not followed the file?

Filing the sheet away after implementation. A sheet archived the day the plan is launched removes the one moment when the organisation verifies that its countermeasures worked.

Grading the sheet instead of questioning it. The mechanism assumes a manager who asks. Where the sheet is used to appraise its author, he writes what will pass, and section II becomes an exercise in presentation.

Where the A3 goes beyond a problem

Shook's examples cover problems, but also proposals, decisions, projects and plans. Verble describes the use at Toyota in strategy deployment, in the multi-level planning of major projects, in cross-functional improvement initiatives and in performance improvement work at the level of a department, a value stream or a unit.

Boundaries with the neighbouring methods

Verble draws the dividing line with 8D at the scope of the problems: the eight disciplines, developed by Ford in the 1980s and run by a team, mainly serve the operational and functional levels, on problems bounded by that team's work; the A3 serves management problems, often cross-functional or strategic. Verble adds that the A3 goes further than most other methods on following up results and closing the PDCA cycle, where 8D stops. DMAIC differs in another way: ASQ defines it as a data-driven Six Sigma project structure in five phases, closed on standard procedures. It is a way of running a project; the A3 fits on a page.

Verble reports that the A3 format was born as a tool for quality circles to report back to their management, before being taught to Toyota's managers to develop their practice of PDCA. Mark Graban observed in 2014 that Japanese quality circles present their work under the name QC Story, most often as a slide deck rather than on a single sheet. The reasoning is close. The single-sheet constraint is the A3's own contribution.

AI considerations

An assistant roughs out section I from notes, minutes and tickets, proposing a business context that the author corrects. It holds the section VI plan: it turns a list of actions decided in the room into a table where each action carries an owner, a deadline and an indicator, then flags the rows where one of the three is missing. It compresses: fitting on one page is a writing problem, and a model is good at cutting a section in half. In a Swiss company whose teams work in French and German, it also produces the language version that is missing, for review by a speaker who knows the domain.

The content of section II stays with human judgement, because it is observed and not deduced. A model asked about the current conditions of a process it has not observed will produce plausible numbers, and an invented current state corrupts everything that follows it, right through to the followup check. The causality of section IV is verified in the field, cause by cause; a model proposes candidate causes, it does not validate them. Then there is the sensitivity of the data: an A3 sheet carries people's names, performance numbers and sometimes customer incidents, which decides the tool it may be entered in.

The mechanism of the technique is the conversation the author holds section by section with the people concerned. A sheet produced in one block by a model, then presented, has removed that mechanism; what is left is a clean document and nothing of what it was meant to teach.

Examples

A Vaud manufacturer of measuring instruments, 180 people, finds that customer orders are slow to be confirmed. The question crosses sales, planning and purchasing, which puts it at the level the A3 is made for.

The links between the sections are read off the sheet. The section III target, 90% of orders confirmed within two working days, picks up the catalogue commitment cited in section I and is measured on the indicator recorded in section II, where 496 orders out of 1'240 run past that limit. The main section V countermeasure, switching on the available-to-promise calculation in the ERP, attacks the cause named at the last link of section IV and not the symptom of the delay. Section VII fixes the two checkpoint dates and names the outcome chosen if the target is not met by 31 March: continue the resolution, because the causal chain is incomplete.

Visualisations

The layout of the sheet carries part of the meaning: the left column establishes what is happening, the right one what is done about it. The arrows say in which order the reasoning is built. A diagram gives that at a glance, a table would lose it. The empty template shows the structure; the filled-in sheet shows the grain of text that a single page allows. The comparison with the neighbouring methods reads line by line and stays a table.

Cost

Tooling

Paper and pencil remain the reference tool: a handwritten sheet is corrected in front of the other person, at the moment when the technique does its work. A fair copy is typed up afterwards.

A word processor or a spreadsheet set to an A3 page in landscape is enough for that fair copy. The spreadsheet holds section VI better, since the plan is already a table of cells. Both have to accept an image, since sections II and IV are shown more than they are written.

An online whiteboard suits a distributed team: the sheet stays one single surface that both parties change during the exchange, which brings the arrangement close to paper. A3-page templates for presentation software circulate widely and work, with a risk of their own: that format invites appendix slides, and the constraint disappears. Whatever the tool, print the sheet at full size before you circulate it. That is the one check that says whether the constraint was held or dodged.

Sources

• Shook, John (2008), Managing to Learn: Using the A3 Management Process to Solve Problems, Gain Agreement, Mentor, and Lead, Lean Enterprise Institute, chapter 1 "What Is an A3?", available as a free excerpt: the definition, the list of the elements of a sheet, the Shook and Verble template, the ten questions that guide the writing and the warning against the idea of a single fixed template.
• Shook, John (2009), Toyota's Secret: The A3 Report, MIT Sloan Management Review, reprint 50408: the same seven-step sequence restated and the statement that the strength of the A3 comes from its size limit and its ordered step structure.
• Verble, David (2020), Dueling Methods: 8D and A3, The Lean Post, Lean Enterprise Institute: the origin of 8D at Ford in the 1980s, the list of the eight disciplines and its containment step, the split of scope between the two methods, the outcomes named at the checkpoint, the depth of reasoning the A3 reaches on following up results and closing PDCA, the reminder that a report describes problem solving without producing it and the origin of the A3 format as a reporting tool for quality circles.
• Graban, Mark (2014), Quality Circles Are Alive and Thriving in Japan, leanblog.org: the field observation of Japanese quality circles, of their reporting under the name QC Story as a slide deck and of how close their reasoning is to the A3's.
• ASQ (American Society for Quality), DMAIC Process: Define, Measure, Analyze, Improve, Control: the definition of DMAIC as a data-driven Six Sigma project structure in five phases, each building on the previous one, closed on standard procedures and process capability.