Rob van Stekelenborg questioned (on Twitter) the Conclusion in my Article “Flows Like a River”
Kanban is Robust, but does it work in Every Case?
Kanban was invented to manage production in a car company. Consider another situation. Your plant makes heavily machined parts for some industrial application with relatively low annual demand. The parts (and there a many different ones) pass through at least 50 operations (assume 60) before they are completed and ready for shipping. The demand for these parts is between 1 and a maximum of 10 per month. A good average demand is 3 per month. Will Kanban work? Work it out for your self before you read on.
Does Kanban Work?
Yes, of course it does. Because demand is so low 1 container with 1 part in it is placed before each work centre (operation). So 60 containers of work in process must be maintained. With an average demand of just 3 per month your plant will have 20 months of work in process lying about. This makes a JIT plant look more like a warehouse with lots of slow moving material. (I doubt any such factory will allow so much material to be in its WIP – there probably isn’t enough space anyway!)
What would Simplified Drum Buffer Rope (DBR) do?
In the majority of factories there is no bottleneck work centre – there is in fact plenty of capacity. We do not have an internal drum – the drum (that gives us the beat for production) is market demand. We also know that generally speaking the market desires short lead times for products they buy from us.
An important piece of information is ‘touch time’ what is the actual processing time of a part relative to the elapsed time from production start through to completion? In almost all factories the number is less than 10% and very often less than 5%. For our example we will be conservative and use 10%. The parts are worked on for a total of 2 months. If production is allowed to finish a part (once it is released into the factory), then it should be out in, say 5 months (it waits in queues for 3 of those 5 months).
Manufacturing follows the following rules:
Start with finished inventory of 12 units (4 months supply)
Every time 1 part is sold a new part is released into production. It will arrive in the warehouse 5 months later.
Since the system has been running for a while, there will be 15 units (5months X 3 units/month) in production somewhere.
There is a good chance that products get to finished product before the 5 months are up (which will cause finished product to increase).
As described the system contains 12 + 15 units – 12 in finished product plus 15 underway in production and possibly in finished product already. That is 27 units (vs. 60 with Kanban)
Since 12 units may be too many production will monitor the finished product stock. If stock levels never approach zero, there is a good case to lower the number of units in the system (similar to eliminating Kanbans)
Using the example, admittedly different from automobile production, which of the 2 methodologies gives the better economic results?
We did not discuss the number of units to be held as finished product in the Kanban system, but which of the two systems can react more quickly to increased (or decreased) customer demand?
Consider Car Production
An assembly line, just like any other production line, must have a work centre that has the least amount of capability. That work centre in fact controls how much the factory output will be.
Kanban factories are famous for giving anyone the authority to stop the entire production line should a defect be identified. That is a great idea because it drives home the need for perfect quality – nobody wants to be the cause of a line stop.
BUT, if the line is stopped the capacity limiting resource also must stop working and that is much more costly than stopping any other resource. The capacity limiting one can never make up for lost time. All other resources can. Conclusion: the capacity limiting resource must never be stopped (unless the defect is fond there.)
TOC and its DBR methodology puts buffers in the appropriate locations – just in front of the constraining resource and an empty space just behind. The buffer in front ensures the constraint does not run out of work during a stoppage – it does not lose capacity. The empty space behind the constraint is there for the units the limiting resource produces while the following machines are stopped. Once the whole line is running again the faster machines before and after the limiting unit fill the buffer and make sure the ‘empty’ space is empty again. The higher capability of the other units’ means they can catch up.
This may not seem like much, but if capacity is an issue the value of an hour at this weakest link is enormous. If a car’s price is 25k€ and 75% of that is materials then every 4 additional cars sold is 25k€ to the bottom line. If the production rate is 32 cars per hour, then 1 hour of constraint time is worth 200k€. Multiply that by the hours gained in a year.
Conclusion
TOC and Kanban are not so far apart – TOC simply takes the Kanban ideas a step further. Place your buffers in the appropriate place!
