Buffer Management is a critical insight, which has ramifications far beyond what we in TOC do right now. The point is that we deal with fluctuations that are NOT described by the well known mathematical distribution functions, like Normal, Beta and Gamma. Buffer management offers a true control mechanism that is not assuming any specific known shape of the fluctuations and even does not rely on the information behind the planning that dictate the size of the buffer.
In order to offer excellent availability, or ensure on-time of an order or a project we include buffers as an integral part of the planning. However, it is impossible to truly determine the “optimal” buffer. Actually I feel inconvenient to use the term “optimal” for managing commitments to the market. We simply have no idea of the true spread of the demand, the supply and the execution. Looking into the past is not good enough because not only we need the combination of the variability of three variables (demand, supply and execution) the three variables are exposed to very considerable possible changes, which change the basic parameters of the variability. For instance, any emergence of a new competitor, a new product or even a marketing campaign has a very significant immediate impact on the demand. The supply depends a lot of the amount of excess capacity and running out of capacity has a huge effect. The level of execution goes through a change every time a new employee starts to work, certainly when a new manager is appointed.
The idea of Buffer Management is twofold:
- Identifying when an order is “almost late“, meaning becoming too vulnerable to additional fluctuations. Based on the red warning we can increase the priority of that order and sometime expedite it by special measures.
- Check the general quality of the buffer by looking on the COMBINED EFFECT of all the three distribution functions! This means using the concepts of “too much red” and “too much green” as warnings to a possible need to change the buffer.
Please note, the main protector of the commitments to the market is still the buffer! The optional warning and expediting are used to add flexibility as an additional mean to protect the commitment to the market.
So, what is the “right” size of the red-zone? Any attempt to calculate the red-zone based on forecasts, capacity and lead-times are damaging without any additional benefit. The sanity check should answer the following questions:
Suppose the red-zone is somewhat too large what is the damage?
Suppose the red-zone is somewhat too small what is the damage?
My assumption is that I’ll never be at the imaginary optimal point. Would you agree that a somewhat larger red-zone, meaning issuing some early warnings a little too early, still representing true priorities, is not all that damaging?
So, the idea of one-third of the buffer is an excellent recommendation, because it is simple and effective and because it has a certain tendency to be somewhat larger than required. The simplicity of the idea, based on our inability to be precise and recognizing the fact that we should separate buffer management from the data used in the planning, looking just into the actual effects, makes it the right way.
The one possible damage of too large red-zone is when too-much-red is frequently observed, but the commitments to the market seem to be under very good control. In my book, Supply Chain Management at Warp Speed (Written together with Bill Dettmer and Wayne Patterson) I give an example for a relatively infrequent situation where one-third is too big. Thus, for that example I recommended 1/6 of the buffer. Calculating the red-zone to be 27% in one case, 23.9% in another and 31.2% in yet another SKU looks to me damaging, not just because of maybe somewhat late warnings, but because the effective logic has been lost!
Something else to think about. Simplified-DBR includes an absolutely required tool of the Planned-Load of one, or more, critical resources. The Planned-Load is a complementary control mechanism to check whether the short-term capacity is enough to support the stock-buffers or the due-dates of the existing open-orders. When the data is good-enough it gives early warning from a situation that might shake the whole system. In other words, the “too-much-red” warning would definitely appear, but then the time to fix the problem might be too short! The Planned-Load is also very effective in determining the “safe-dates” for make-to-order, so it is both an execution and planning device. To my mind, the Planned-Load is an integral and vital element of ensuring stable fulfillment of the company commitments.
7 thoughts on “Why should the Red-Zone be 1/3 of the buffer?”
Thanks you so much Eli for sharing this insight. Very important.
I do agree to keep it simple as 1/3 of the buffer to be Red, What according to you would be best percentage distribution of Black,Red,Yellow and Green. My experience in case of MTA scenario is whenever we start focusing more on Red production starts building protective capacity in terms increasing the size of buffer and there by ending up with higher percentage of Green.
Note: Decreasing the buffer using DBM becomes difficult task as most of the attributes are assigned to bulk sales
The realistic conflict is between too short buffers, sometimes suffering from shortages, and too much stock – relatively long times in Green.
Generally speaking we like to have zero blacks for items that are on availability commitment. Red might be anything up to 10% of the items, and if more I’d start to worry, but there are situations where 15-20% in Red still feels pretty much under control.
If you know the bulk sales in advance, I’d suggest to treat such orders as MTO.
nice post, as usual!
May I ask you what software would you suggest to manage a shop floor the S-DBR way?
Thank you for your attention
There are several S-DBR packages around. My son, Amir, manages a software company called Inherent Simplicity that developed a packaged called Symphony. As said, Amir is my son, so I’m objective. However, the software was influenced by me to a large degree (another reason I’m not objective on this matter) and closely monitored by Eli Goldratt.
Thank you for your kind answer. I’ll try to contact them to check if there’s free trial available and the relative pricing 🙂
The decisions made on the buffers affects the capacity constrained resource. As the mechanism to release considers the size of buffers and the current planned load, should we consider the location of the CCR as part of the algorithm to sustain the reliability of the system? I’m also assuming that BM is in place.
I think you have write something about this in the past, can you please share a link with this insights?