Toyota Production System

It has had 55 responses to date, to which I added my own this morning:

In 1986, I was trained by Eli Goldratt’s organization to implement their OPT production scheduling software, which was the first incarnation of TOC. My first project afterwards was to implement it in an aluminum foundry that produced castings for the auto industry, had 40% of its diecasting machines down at any time, and took 40 hours to change a dies in any of these machines. It was obvious that scheduling was the least of their worries, but it was what corporate management had decided to focus on.

As OPT was not a viable software product, Goldratt’s organization soon dropped it to focus on the “thoughtware” behind it, repackaged as TOC. I read The Goal at the time, and, later, Theory of Constraints, and found these theories simplistic and misleading. Simplistic because they fail to account for the variety of challenges facing manufacturing operations; misleading, because they make managers believe that you can make all the improvements you need without moving one machine or doing any engineering improvement. One point I remember from The Goal was focusing “setup men” on the bottleneck machine; the idea of improving setups so that they could be executed by production operators was not considered.

Most of the above comments, about balancing, heijunka, and kanban, are about assembly, and one key characteristic of assembly work is that it is almost infinitely divisible. You can break it down into elements that are as small as you want and reassign them between stations backwards and forwards. This is why an assembly line does not have bottlenecks, and a book on assembly does not need the word.

Machining and fabrication are a different story, because there are many cases where it is technically impossible to reassign work between machines. There are also enormous differences in cost between machines, for example between a drill press and a gear hobbing machine. As a consequence, a gear cutting cell has a bottleneck machine that you try to keep as busy as possible, and supporting machines that you try to make sure are available when needed but don’t care if they are used only 10% of the time.

Capacity planning and production scheduling in such a cell is centered on its bottleneck machine, but. on the scale of a cell, it is not complex. If the cell is dedicated to a product, it is trivial; if not, the key challenge is to achieve quick changeover for the entire cell so that it can support heijunka for the entire plant.

TOC is a one-trick pony. TPS is a system that addresses the full variety of manufacturing challenges found in the automotive supply chain, and it has been successfully adapted to other contexts with the “Lean” label.