We recently received an e-mail note from someone who had attended one of our seminars, with this question:
“We basically ran into a wall attempting to implement a Pull/Kanban solution because the Vice President of Manufacturing is concerned that although the Pull solution seems to be a better fit for the (repetitive automotive parts) it would not work for the other product lines and work centers. Some of these (injection molding and painting) are common for all product lines . . . You had a slide in your material under "Multi-Mode Mfg. Summary” that suggested you can operate a part of your plant in Lean mode and yet keep a portion in traditional ERP. I am very interested in learning more about this.”
This line of questioning comes up often and begs a few fundamental questions. Before we discuss them, let’s review some of the key considerations in the implementation of Pull systems.
First Things First
For our purposes, a Pull system is any form of consumption-based trigger for replenishment or “Replenishment that is triggered by consumption (as opposed to planned and “pushed” to point of use).” The term Kanban refers to types of signals for replenishment. Translated from Japanese, Kanban loosely translates as “signboard” or “signal card”.
Pull signals can take literally a thousand forms, like removable and reusable cards, dedicated containers, empty spots on the floor, colored golf balls, red lights, sounding beepers . . . and just about anything in-between.
Sounds pretty simple, right? Maybe everything can be under Pull/Kanban control? Maybe not.
Demand Patterns Dictate Pull Signal Success
If average demand over a given horizon is relatively stable or predictable (say, plus or minus 10% in daily/weekly usage), simple and manually executed Pull signals will work very well.
If demand varies wildly from period to period, say 100 this week, 10 in week two, and 250 in week three, we suggest re-thinking the situation. First, is the demand variation driven by customers or is it internally driven? If it is internally driven the first steps may be working to level out the schedules to permit Pull signal effectiveness.
For High Demand Variation Items There are More Questions
If demand variation is customer driven and it is impossible to negotiate a more level demand pattern, things get more difficult. This is where some decisions are needed – should the item be managed with conventional MRP (Material Requirements Planning) logic, or is there still a good business case to apply a Pull signal? Some things to consider are lead times, relative cost, obsolescence and shelf life issues.
Let’s examine each of these in turn:
- Lead Times – if lead times are relatively short, Pull signals may still make sense, depending on the other factors we are about to discuss. If lead times are excessive, there may be little alternative to conventional planning and control scheduling methods.
- Relative Cost – if this is a low value, “C” level item that is relatively small, the attractiveness of using Pull signals increases dramatically. The average cost of tracking, ordering, receiving and counting an item can be $200 or more per transaction. For an item with annual consumption under $1,000 in value, it will be very attractive to use Pull signals. On the other hand, very expensive “A” inventory items may need to be treated differently, as the benefit/tradeoff of Pull signals may not make sense. If you are forced to put three months of inventory in for an item to cover variation in the Pull system, you will likely be much better off using conventional scheduling and control methods to bring these items in daily, weekly or monthly.
- Obsolescence Risk – if the relative design stability of this item is good, with three years or more expected before changes, then Pull signals will not be overly risky. Remember this: Pull systems are NOT Zero-Inventory Models. Inventory to support the replenishment trigger cycle will always be in place. High “risk of design change” items should be managed very tightly – again with a conventional scheduling approach.
- Shelf Life Issues – finally, if there is wild variation in demand AND shelf life issues, Pull signals may not be the way to go. As with obsolescence risk, we must remember that Pull systems are not Zero-Inventory Models. The larger amount of buffer inventory that has to be loaded into the Pull signal quantities will likely result in some stale goods . . . a bad thing.
Getting Back to the Original Question
What we have just described is the basis for a “Multi-Mode Manufacturing” approach to managing items in inventory – using Pull systems on some items and conventional scheduling management approaches on others in the same facility.
For our case question at the beginning of this article, we would need to find out what the variation in demand patterns look like on an item-by-item basis. If you look one more level down from the finished level, maybe we will find common elements, such as common raw stock used on many different parent parts. We might find that total demand in pounds and hours of capacity is relatively predictable.
If so, the basic components can easily be under Pull system control and the actual final value-add operations can be driven directly by customer orders. In this case, you end up with a hybrid (multi-mode) execution environment, stocking materials and making basic common sub-components under Pull system control while performing final “make” operations on a “make to order” tightly controlled basis.