How to write an A3

An A3 is a structured way to work one problem on one sheet of paper. The name comes from a paper size close to 11 by 17 inches. The sheet limit is the point. If you can't fit your thinking on one page, you probably haven't finished thinking.

Published A3 formats vary. Many use five parts: problem, current state, target condition, root cause and action plan. This article uses an eight-step version that splits those parts more finely. It closely follows the sequence often published as Toyota Business Practices. Each step asks one question, and you don't move on until you can answer it. You can write along in the free A3 builder. Select Load sample to see the example from this article.

The example is fictional. It comes from a made-up metal parts plant, and its numbers show the method. They aren't results.

Before you start

An A3 needs three things:

• One problem. If your title has an "and" in it, you may have two A3s.
• One named owner. That person writes the A3. Their supervisor reads it, questions it and signs off on it, but doesn't write it.
• A goal it serves. Link the A3 to an objective in the plan. If you can't, ask why you are spending time on it. Improvement work that isn't tied to an objective can succeed and still not move the strategy.

The eight steps

Here is each step, with what the second-shift supervisor at the fictional plant wrote.

1. Define the problem and the gap

Say where things are now, where they need to be and by when. Use a number.

Line 2 scraps 48 parts a week. We want 24 or fewer a week by March 31.

Don't name a cause or a fix yet. "We need better setup sheets" isn't a problem statement.

2. Break it down with data

Break the number into counts by time, job, place and likely cause. Go and look, and count what you see.

Of the 48, 31 are bend-angle defects. Of those 31, 26 happen in the first hour after a changeover. The other 17 are scratches and wrong blanks, with no pattern yet.

This is the step that people most want to skip, and it's the step that changes the most minds. The supervisor might have guessed at operator skill. The count points at changeovers.

3. Set a target against the opportunities

Using the data, say how much of the gap you can take out.

First-hour bend-angle defects are 26 of the 48. The target is to take out those 26, which leaves 22 a week.

Now the target has a reason behind it. It's no longer just a number that someone handed down.

4. Pick the priority, find the point of occurrence and ask why

Pick one piece of the problem that is within your authority. Say exactly where it happens. Then ask why five times.

Point of occurrence: Line 2 press brake, first hour after a changeover.

1. The back gauge and ram depth don't match the job. Why?
2. The operator sets them from memory of the last run. Why?
3. The setup sheet has no gauge or depth values. Why?
4. The sheets were written for the old machine, and nobody updated them. Why?
5. Nobody owns the setup sheets after a machine changes.

Five is a guide, not a rule. Stop when you reach a cause that you can fix. If your last why is "people don't care", go back. You've stopped looking at the process.

5. List countermeasures and rank them

List several countermeasures. For each one, ask: does this remove a cause, or only hide it?

The supervisor starts with the setup sheet, and uses the first-piece check as a stop-gap until the sheets are ready.

6. Plan who does what, by when

Name a person and a date for each action. Trace each action back to a cause. Say when you expect to see results.

Note the third row. It fixes why 5, the deepest cause, and it needs someone outside the supervisor's own team. The supervisor raises that in a one-on-one, along with the back gauge purchase.

7. Check the process and the results

Set check dates in advance. On each date, ask two separate questions: are people doing it, and is the number moving?

On January 29 and February 26, ask two questions. Are operators using the sheet at each changeover? Is the weekly count moving? If they use the sheet and the count doesn't move, go back to step 4 or 5.

The two answers tell you different things. If people aren't doing it, the fix isn't in place yet. If they're doing it and the number doesn't move, your theory was wrong. Each case needs a different response, so keep them apart.

8. Standardize and share

Lock in what worked, so that it survives the next new hire. Tell the people who should know. Then look for the next place that the problem happens.

Add the setup sheet to new-operator training. Supervisors check for it at each changeover. Share the sheet with first shift and Line 3. Then go back to step 4: where else does this happen?

The loop back to step 4 matters. The first A3 went after one point of occurrence. The goal may still have a gap, and the remaining 17 scratches and wrong blanks are still on the sheet.

Here is the whole A3 on one sheet.

How a supervisor should read an A3

If you sign off on someone's A3, your job is to question it, not to rewrite it.

• Check the logic from one step to the next. Does the target follow from the data? Does each action trace to a cause?
• Ask what the person saw, not what they think. "Where did you count that?" is a useful question.
• Don't sign a step that you don't believe. Ask for more data, and let the owner go and get it.
• Don't supply the answer. If you write the fix, the person learns nothing, and the A3 becomes yours.

How Run Hoshin handles this

The Run Hoshin app uses the same eight steps. An objective arrives as step 1 of the A3 for the person who holds it. Each step opens only after the supervisor signs off the step before it, and you can't sign off your own step. After step 8, the A3 loops back to step 4 for the next point of occurrence, against the same goal. The app keeps a history of each step. On screen, an A3 is a stack of steps, not one sheet, so the sheet limit is yours to keep. The app has no one-page print view and no file attachments, charts or photos, so you type your data into the steps. You rank countermeasures in words, and the app computes no risk score. The free A3 builder prints, but the app can't import its file. See what the app does not do yet.

Sources

• Lean Enterprise Institute, A3 Report