I’m not talking about the smoke and mirrors of a Las Vegas act. I’m talking about real magic where parts and materials suddenly appear at the right quantity, the right place and at the right time. I’m talking about the two-bin system, reorder points and statistical forecast models.
It has been said that Henry Ford used the two-bin system in the early production of his automobiles. As you know, each bin had to be equal in size to hold the quantity used during the lead time of the part with a little extra for unforeseen emergencies. When the first bin was emptied, a reorder card was sent to Purchasing to reorder another bins worth.
The two-bin system assumed, falsely, that all demand quantity was, for the most part, steady and continuous. Because of this false assumption, there were many problems with stock shortages and inflated inventories. There was also a problem where you had to stock lots and lots of different sized bins to operate such a system. In the 1930’s some mathematicians worked to eliminate the “lots of different bins” problem by devising the Reorder Point quantity formula.
Here they said the reorder point quantity on hand was equal to the demand during lead time plus some added safety stock (RP = DLT + SS). This formula allowed the two-bin system to still be used and stocked on any shelf without the bin problem but still falsely assumed all demand to be steady and continuous while still generating stock outs and inflating inventory levels.
Can you see the magic in the two-bin Reorder Point formula? A lot of people can. Reorder points are still used in many of the new sophisticated ERP software packages today. Heck, I even read a published article where one author thought that MRP didn’t work for whatever reason and suggested we should do away with MRP in favor of a 3-Bin Kanban with two bins on the shop floor and one at the supplier.
After reading this I remembered the old saying that “Those ignorant of history are doomed to repeat it”. By now you can figure out that I don’t believe in two-bin, reorder point magic, however I do believe in statistical forecast models. I believe that all forecast models are wrong.
I don’t like to forecast at all but I realize that sometimes you just have to. I’ve been trying to come up with a forecast model for the six numbers of the Powerball drawing for years. When I come up with an accurate model, I’ll let you know. Forecasts are simply guesses requiring constant vigilance and adjustments to history. Two-bin reorder points assume parts are totally independent of each other so you have to forecast for every single part. That’s a lot of guessing and adjustments to history which requires a lot of maintenance.
The MRP planning process was devised to avoid most of the smoke and mirrors hocus pocus by a dramatic reduction in the amount of forecasting (guessing) you have to do. MRP has product structure. The idea here is only to guess where you have to (Independent Demand) and derive (Dependent) demand from those fewer guesses via a product structure therefore requiring much less maintenance. When you need reorder points for “C” items use the MRP generated Time Phased Order Point (TPOP) which utilizes independent, derived, unsteady and discontinuous demand, a future blog topic of mine.
Discussions
I found your article very amusing because when I was working for a contract manufacturer, this EXACT thing happened to us. Our customer wanted to 'TURN OFF MRP' and drive a replenishment system down to every component in the supply chain. They wanted to make their supply chain lean.
Even though many of us in the planning area were adamantly against this approach, being a contract manufacturer, we did what all self-respecting contract manufacturer would do in the electronics industry, we said, 'yes'.
It was a NIGHTMARE! Rather than focusing on the 1000 assemblies and continuing to adjust and tweak their demand, you multiplied that by 100 and you now had to adjust the buckets on 100,000 components. If their was a black-swan drop-in . . . there was always a black-swan drop-in . . . then you had to employ single-use kanbans (SUK) on numerous components. The acronym speaks for itself. To counter-act the effect, safety stock levels had to be driven through the roof. All parties quickly realized they had a problem. Ultimately, they had to create a 'parallel system' to start adjusting these replenishment orders according to MRP. Politically none of the parties would admit they went back to MRP and the component level, so they still had to maintain additional overhead to keep both systems running in parallel.
The overall project did have some very good positives. It drove the customer to focus on better forecasting and reconciling gross deviations to their historical rate. It also re-aligned the contract manufacturer to replenish to real demand verses a forecast created 2 months ago at the start of the quarter.
So when do replenishment systems work well? They work well when they drive immediate shipment or manufacturing triggers. For manufacturing lead-times, they need to be relatively short and the manufacturing facility needs to have a consistent throughput time. Finally, the management team realizes that when a black-swan event occurs (and they always occur) that they are going to have to wait.
For most, if not all, industries, MRP still needs to be the backbone, but replenishment (2-Bin, ROP, Kanban) can be applied STRATEGICALLY as a downstream advancement trigger.
Thanks for your feedback. I knew there had to be someone else out there that would have had to be scarred by main assumption of two-bin or reorder point replenishments, being that all demand is steady and continuous. It is NOT and that is a fact! Why do we go on believing in this magic?
You made an interesting point of safety stock going througn the roof which is typical of these replenishment systems. I'm writing a blog right now on my experience with a large public utility here in Arizona that used computerized reorder points for the scheduling of their construciton materials and parts. Do you wonder why your electric bill is so high?
Yes, maybe you can use ROP but with very short lead times which doesn't happen very often.
Thanks for your comments and stay tuned.
Ray
Leave a Reply