Ohno on waste and pull
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
🏛️ Taiichi Ohno wrote a short, stubborn book about one idea: make only what the next step has taken. Everything else follows from it. A factory that forecasts and pushes will always make too much of something, and he ranks that above every other waste, because overproduction hides the rest. Stock covers for the broken machine, the bad part, the late supplier. Take the stock away and the problems stand in the open, which is the point.
The two supports he names are just-in-time and what gets translated as autonomation: a machine, or a person, that stops when something is wrong instead of producing defects at speed. The signal that carries the pull is the 🔤 Kanban, a card that travels back up the line and authorises exactly one more batch. He says he took the idea from the American supermarket, where the shelf is refilled because a customer emptied it.
The method for a problem is to ask why five times, until the answer is a cause and not a culprit. That is the part I already believed. The part that changed my mind was the order of his wastes.
✨KEY IDEAS
Making more than the next step needs is the worst waste, because it buries every other problem under inventory.
The later process goes to the earlier one and takes what it needs; the earlier one makes only the replacement.
Any worker can stop the line, and a stopped line is treated as a finding.
Ask why five times and the fifth answer is usually a rule or a missing check, not a person.
He designed for slow growth and many variants, where large batches stop being cheap.
📝READING NOTES
Paraphrase, not quotation. Read in bed, 13 to 18 October, signed off sick for the first three days.
The argument, in order
- Start from the customer and count backwards. The final step knows what was sold. Each earlier step only needs to know what the next one took.
- Small batches, levelled. If the last step takes evenly, the steps before it are not whipped about.
- Make problems visible. Lower the stock until something breaks, fix that, lower it again.
- Stop on a defect. Never pass a problem forward to be found later at higher cost.
- Find the cause. Five whys.
Where I disagree
He writes as if every waste can be seen from the floor. In software the inventory is invisible: half-built features, a queue of unreviewed work, a list of open exceptions. Nobody trips over it. My dispatchers at the pilot had 214 open items each before nine and no shelf to show it.
He is also harder on slack than I am. A system with no buffer is brittle, and my newsletter is named after the opposite instinct. I think the reconciliation is that he removes stock and keeps capacity: his people stop and help. That is slack, held as time instead of as parts.
What I took
- 💎 Local efficiency can lower the output of the whole: a team that is always producing is making inventory for someone. Same finding as 📖 Goldratt on local optima, reached from the other end of the factory.
- 💎 An exception is information about the process and not a nuisance in it: the stopped line.
- 💎 Pull beats push when the downstream step knows more: we push every exception at the dispatcher. They would rather ask for five.
- Concept written up as 🧩 Pull systems.
For the scorecards work: a carrier score that nobody downstream asked for is overproduction with a chart on it.
🗣️IN SHORT
- In : Toyota Production System: Beyond Large-Scale Production, Taiichi Ohno, English edition 1988
- Ohno argues that : production should be pulled by what the next step has actually used, and that making anything beyond that is the worst waste
- by stating that : excess stock hides defects, breakdowns and bad plans, while a card passed upstream and a line that stops on a fault bring them into view where they can be traced to a cause