Hansei: the honest reflection that holds up Kaizen
At Toyota they do hansei even when the project went well. That's the whole difference with our closing meetings.
Hansei translates as "reflection", but in Toyota culture it means something considerably more demanding: honestly acknowledging what didn't go well, without hunting for someone to blame and without dressing it up. And doing it before any improvement, not after.
Enter your email and you will get an alert when something new is published. Unsubscribe anytime.
It's the counterpart to kaizen. Kaizen looks forward: what do we improve now? Hansei looks back: what actually failed, and why?
Without the second, the first stays a list of good intentions.
What separates it from a closing meeting
And here comes the detail that throws everyone who sees it for the first time: at Toyota they do hansei when the project went well too.
If the result was good but you were late, had to improvise, or the team ended up wrecked, that goes on the table just the same.
Our closing meetings work exactly the other way round. If the final number is good, there's nothing to review and on to the next thing. The result? You consolidate ways of working that succeeded by luck.
| Usual closing meeting | Hansei |
|---|---|
| Happens when something goes wrong | Happens always, however it went |
| Looks for the person responsible | Looks for the system's cause |
| A good result closes the subject | A good result gets examined too |
| Ends in conclusions | Ends in a concrete action |
Why it's so hard
Because it asks for a cultural condition that is extremely difficult: that admitting a fault doesn't cost you personally.
In an organization where pointing at your own mistake gets charged to you, hansei is impossible however many meetings you call. People will learn — and they learn fast — to present things in a way that leaves nothing to reflect on.
And that's the real reason lean fails in so many companies that copied its tools. The boards and the audits get copied. What doesn't get copied is the safety to say "I got this one wrong".
The bridge to 2026
From here I take something very practical for artificial intelligence projects. And it runs against how they're normally sold.
An AI system in your plant gets things wrong. It flags a defect that wasn't one, or lets one through that was. It's going to happen.
What decides whether the project survives isn't the accuracy on day one. It's whether there's a circuit where the operator can say "this thing you've flagged is fine" without a row starting, and whether that correction goes back into the system so it learns from it.
That's hansei applied to a machine: acknowledge the fault, without drama, and turn it into an improvement.
The projects I've watched run aground didn't fail because the model was bad. They failed because nobody designed that circuit, and the operator ended up ignoring the alerts in silence. Which is the worst way to lose a project, because you don't find out until very late.