Hansei: how to run a reflection that changes the next job

Five things separate hansei from a vague look back:

Scale it to the work. A small task may need three lines in a notebook. A large project may need records, other people's views and a longer conversation.

Hansei pairs honesty with a sense of scale. Replace a judgment about character with an observation you can act on:

The right-hand versions point to a decision or a gap in the process, so they suggest something to change. The left-hand versions only assign a feeling.

Sort the causes into three rings before choosing what to change:

Suppose a supplier ships late. You did not control the supplier, but you can ask whether you checked the dependency, flagged the risk early or had a fallback. The commitment might be earlier escalation, not a promise that delays will never happen again.

In a team reflection, apply the same rings to management decisions: priorities, staffing, approval steps and access to information. Don't ask one person to keep covering for a process the organization is unwilling to change.

The practices overlap but produce different things:

Hansei is sometimes compared with the Check step of PDCA, because both stop to examine results before moving on. Reflection finds what needs attention, kaizen tests a change, and an updated standard work keeps what worked. Calling a meeting hansei does not make it one; examining your own part and committing to a change does.

One practical routine, from intent to shared lesson:

1. 1

   State the intended outcome. Write what you were trying to achieve: quality, timing, effort or communication. If expectations were unclear or in conflict, record that instead of inventing a standard after the fact.

2. 2

   Describe what actually happened. A short, factual account built from the work itself, messages, decisions and handoffs. Keep observation apart from judgment, and note what went well too.

3. 3

   Examine your part and the conditions. Ask what you assumed, what you noticed and didn't act on, and what in the roles, workload, tools or approvals made the planned action hard. Do both; don't pick one.

4. 4

   Choose a proportionate countermeasure. Turn the lesson into a change you can see, sized to the cause. Name an owner, a finish point and the result you expect, and think about side effects such as a new review creating a queue.

5. 5

   Test it and review the result. Use the change on the next real task. Decide beforehand what evidence would show it worked, and check what people actually did, not only that a step was added.

6. 6

   Keep and share what works. Update the instruction, template, handoff agreement or training where the work happens. Toyota calls spreading a lesson to similar areas yokoten. Share the reason with the action so another team can judge whether it fits.

If the change fails at step 5, go back to step 3. The cause may have been incomplete, or the countermeasure impractical.

Use these when a reason sounds vague:

• What did I assume without checking? For example, that an earlier conversation counted as sign-off.
• Which warning did I notice but not act on? A late reply, a missing file, a question nobody answered.
• Where was ownership unclear? Two people each thought the other had it.
• What made the planned action hard? Workload, tools, timing or a missing decision.
• Which explanation still needs evidence? Mark it as a guess and say how you will check it.

Keep asking for a more exact account. If the cause is poor communication, name the missing information, the sender, the receiver and the point in the flow. Stop when you have an action you can take, not when someone has accepted the blame.

Reflect at points where the lesson can still shape the next choice:

Deal with an urgent problem first, then reflect with the facts and people you need. And stop when you have a lesson, a next step and a review date. More discussion past that point usually adds little.

Four illustrative cases, each ending in one testable change:

Fill it in plainly and keep only what supports a decision. Where you lack evidence, write not yet known and how you will find out:

Notice what the entry leaves out: a judgment about personality, a promise never to make mistakes and a long list of extra improvements.

Agree the purpose before people arrive: which event, what information to bring and which decisions the group can make. If some changes need approval from outside the room, say so up front.

Start with your own decision. For example: I added a late priority change without checking its effect on the handoff. Then ask questions, rather than waiting for everyone else to confess. Use simple ground rules:

• Describe actions, not identities: A missed check, not a careless person.
• Invite the people who did the work: And the people who received it, not only those who managed it.
• Separate disagreement from misconduct: Different accounts of the same event are normal.
• Keep explanations open: Say what the evidence supports and what still needs checking.
• Finish with named owners: Every action has a person and a date, not the team.

If the talk turns defensive, return to one specific event and what was known at the time. Ask what made the choice seem reasonable then. Before closing, read the actions aloud and check that each owner has the authority, time and resources to do it. Reflection meetings are one use of retrospective tools; kaizen and A3 problem solving take over when a gap needs a full improvement project.

Check your format against these:

1. 1

   Reflection means finding who is at fault. Look at personal choices and process conditions together. Accountability names a useful responsibility; it does not end the inquiry.

2. 2

   Only failures need review. Small gaps and successful work are worth examining, especially when success depended on extra effort.

3. 3

   More criticism means deeper reflection. Depth comes from specific observations and sound reasons, not a longer list of negatives.

4. 4

   An action item closes it. Plan to check that the action happened and that it helped.

5. 5

   A template guarantees honesty. A form supports the inquiry; it does not replace it.

6. 6

   Everything must change at once. Pick the most important gap and test one change.

A final check: read your conclusion on its own. If it says only improve communication, take ownership or try harder, make it specific. Say which behavior or setup will change and what you will look at to know.