Christoph Roser has more impressive credentials than most Lean consultants, from a PhD in Engineering to a research job at Toyota labs, stints in operations at Bosch, and a professorship at Karslruhe University of Applied Sciences. So, if anyone is qualified to write a theory of Lean, he is, and he is trying his hand at it in production planning and scheduling.
It is an effort that should be encouraged. As Takahiro Fujimoto put it, the Toyota Production System (TPS) emerged from a company’s practices, not from a of theoretical vision. Nicola Tesla designed the AC motor in detail before AC power was available, and Alan Turing designed the computer years before the first one was built, but nobody designed TPS before the first car was built. It was instead an incremental development, based on the solutions Toyota people used to overcome a succession of crises in the company’s growth.
As a result, to this day, the theory of TPS, or Lean, is weak and half-baked; I have written on this before. On the subject of Leveling/Heijunka, Roser has, to date, posted the following.
- Why do Leveling (Heijunka)? I don’t see Roser’s post as truly answering the question. The purpose of Heijunka as smoothing the work upstream from the operation you are doing it on. You want your feeder lines and your Tier 1 suppliers to have the smoothest possible work load, given the variability of your own. If they do the same for their own feeders and suppliers, you defeat the bullwhip effect, in which small demand fluctuations at final assembly are amplified up the supply chain into alternations of feast and famine. American and European managers have been slow to take up this challenge because their education has taught them to worry exclusively about their customers and not their suppliers.
- An Introduction to Capacity Leveling. The vocabulary is a bit confusing. His recommendation to “Keep Capacity Constant,” on its face, is obvious, because you don’t buy and sell machines on a daily basis. What he really means is “Keep Capacity Utilization Constant,” which is a different story.
- Theory of Every Part Every Interval (EPEI) Leveling, Also Known as Heijunka. I am used to thinking of the EPEI as a metric, the “Every-Part-Every” Interval, meaning the time it takes for a production line to cycle through all the products it makes, and it is a time that you try to shorten in order to reduce your order fulfillment lead time as well as your finished goods inventory. What Roser describes here is the idea of rotating production through a fixed sequence of products in fixed quantities every period.One key motivation for a fixed sequence that he does not mention is having setup times that depend on both the “from” and the “to” product. With a from/to matrix of setup times, there is a sequence that minimizes the total setup time needed to cycle through all products that you should follow. It does not mean, however, that the quantities you make of each product should always be the same.
- The Folly of EPEI Leveling in Practice – Part 1 and Part 2 The following is what I most agree with in all of Roser’s posts: “Lean is not a religion. Lean is hard work, and you actually need to understand what you are doing. Just copying something without understanding is a good way to fail, especially with leveling.” As reasons for the failures he has seen in EPEI leveling, he lists the following:
- Inability to Closely Follow a Production Schedule
- Lack of a Structured System to Handle Changes
- Inability to Have a Good Prediction of Customer Demand
- Lack of Additional Inventory to Cover Increased Replenishment Time
- Limited Power Over Supplier and Supplier Ability
- Does Management Want to Be Cheated?
These issues are not particular to EPEI leveling. In fact, most advance planning and scheduling system (APS) fail because production cannot execute the instructions they issue. As to the question of whether management wants to be cheated, the answer is clearly yes. Otherwise, management would not demand unrealistic sales forecasts, or pressure project managers to commit to impossible deadlines. Roser’s recommendations to “Limit the Damage of EPEI Leveling” strike me as simply methods to make it work.
- Introduction to One-Piece Flow Leveling – Part 1 Theory, and Part 2 Implementation. The sequencing algorithm Roser describes is the simple one, that only works to smooth the upstream flow if the products have no common parts. Otherwise, you have to use the more complex approach described in text in Monden’s Toyota Production System and graphically in Lean Logistics to choose the sequence of products so as to smooth the incoming flow of parts.In his discussion of “When Lot Size One May Actually Be Too Small,” Roser does not discuss customer order quantities. If your customers buy screws by the box, the product that you sell is a box of screws, not an individual screw. It is difficult to come up with a reason to alternate production among screws one unit at a time when what you are making is boxes.
I would certainly appreciate Roser’s feedback about Chapters 13 to 17 of Lean Logistics, which covers the same ground, with more emphasis on the analysis of demand patterns as a foundation, particularly the breakdown among products and parts between Runners, Repeaters and Strangers, and seasonal or cyclical variations.