Bill Waddell, intellectual sparring partner for almost 20 years now, has put out this video revealing “The Truth About Kanban”:
Michel Baudin‘s comments:
This video is just Bill’s talking head against the background of a brick fireplace with a few books on top, notably “Toyota Kata.” It contains no moving pictures of Kanbans in action and all you learn from viewing is in Bill’s words, and I have a few quibbles with these words.
I usually get impatient with this kind of video, because voice is a slow medium, and you would get the same information five times faster reading the transcript. But I have never met Bill in the flesh, and I was curious to hear his voice. It’s a good radio voice, albeit curmudgeonly, reminiscent of a younger Tommy Lee Jones.
Now, about the content, Bill makes three main points:
- What you use for a pull signal doesn’t matter.
- You can use Kanbans with long lead time items.
- The Kanban system is a mechanism to drive improvement.
I agree with Point 3, but find Points 1 and 2 problematic.
In Point 1, Bill asserts that it makes no difference whether, “the information is on a card, embedded in a container, on a computer screen, in lights, or people just talking to each other.” It is like saying that using coins, bills, checks, or credit cards makes no difference because it’s all money. In practice, I don’t think many of us would gladly revert to carrying around purses of coins.
In my experience of the Kanban system, the choice of pull signal can make the difference between implementation success and failure. Even at Toyota, it doesn’t have to be a card, which in 1950s information technology; they have been using eKanbans for 20 years. The different types of pull signals support different replenishment protocols that must be sophisticated enough to do the job, yet easy enough for the humans involved to understand, and a simplification of their work. This is explained in Chapter 10 to 13 of Lean Logistics.
The term “demand pull” may be misleading, because, in the Kanban system, the trigger is not that a unit has been sold, but consumed. Toyota’s Kanban system does not trigger the production of a car whenever one has been sold, and dealers hold new cars that have yet to be sold. It comes into play after dealer orders have been processed into a master production schedule and items in that schedule sequenced through heijunka, with the goal of smoothing the consumption rates of components and subassemblies. The Kanban system then kicks in to regulate the replenishment of components in response to the remaining fluctuations.
Bill’s Point 2 about long lead time items is puzzling. He asserts that using Kanbans for them is better than using forecasts but, using Kanbans to replenish an item is implicitly making the naive forecast that demand in the immediate future will equal consumption in the recent past. Naive forecasting is difficult to improve on a time scale of hours for products made at takt times in a few minutes, but the quantity you consumed in the past 30 minutes doesn’t tell you much about what you will need 4 months from now, which is a common lead time for auto parts procured from overseas.
An approach for dealing with such long lead time items is to have a logistics company operate a consolidation center near the plant and have the plant use Kanbans to order these parts from the consolidation center as a local supplier. It doesn’t solve the problem of managing the supply of long lead time items for the whole supply chain, but it gets it off the list of issues the manufacturing plant deals with.
Even with local suppliers, the Kanban system is not used instead of forecasting but as a complement to it. Rolling forecasts are issued to suppliers for information — sometimes as part of agreements that provide compensation for repeated errors — and these forecasts are for information only, with actual orders coming in the form of Kanbans. (See Chapter 17 of Lean Logistics.)