Monitoring the Target Level Size-Dynamic Buffer Management

While the immediate process of fast replenishment of sales and following the right priorities in the shop floor have been covered, the next step is getting the right feedback to the planning stage. The most obvious planning decision is the determination of the target levels. The first initial estimation might not be adequate or changes in either the demand or the supply may have made a certain target level no longer adequate. What should the signals be that a specific target level is too high or too low? It certainly must be shown in the behavior of the on-hand stock. The algorithms for recommending changing the target levels are based on certain behavior patterns of the finished-goods stock and are called Dynamic Buffer Management (DBM).

Too Much Green-the Target Is Too High

When we hold target levels of several items too high, there are obvious negative effects. The direct financial implications and the risk of losing on investment are probably not too prominent assuming the initial determination of the level is not vastly wrong. However, holding too high inventory means we replenish when there is no real need. Therefore, it has a direct capacity impact that at off-peak time might not be problematic, but at peak times could be critical.

The most obvious signal that a buffer is too large is that it is too often and too long "in the green zone." Stock buffers are not supposed to be in the green for too long. It means that the relationship between supply and demand does not call for such a large buffer. We call the situation "too much green"-a signal that the buffer target is too high.

Let's define a parameter called the "green check period" that whenever an item spends time continuously in the green it is recommended that the target level be decreased. The recommended default time for the green check period is twice the replenishment time. The point is to be reasonably conservative. It is not desired to reduce the buffer (the target level) and after a short time to increase it again. Frequently, holding a little too much inventory is preferred to holding too little.

Once the target level is reduced, it is natural for the current on-hand stock to be above the new target level. No checking and definitely no decision on further reduction of the target should be considered when the current on-hand stock is above the top of the green, which is equal to the target level.

Once it is decided to reduce the buffer, the next obvious question is by how much? Goldratt suggests (Strategy and Tactic tree MTS to MTA 2008, entry 5.112.1) to reduce the target level by 33 percent. The topic of by how much to increase or decrease buffers and when to refrain from doing so is a worthy topic for discussion. We'll deal with it later in this chapter.

Too Much Red-the Target Is Too Low

"Too much green" is the signal that the buffer target is too high, and too much red points to a buffer target that is too low. That said, we would like to be a little more precise about increasing the buffer. Spending a lot of time near the top of the red looks bad, but may not be bad enough to propose increasing the buffer. In addition, it could be that every time the buffer turns red it is replenished quickly, but soon turns red again. This might be a signal that the buffer is not high enough to prevent the risk of shortages.

The idea is that both the amount of time an item spends in the red and the depth into the red are relevant signals to increase the buffer. The algorithm that emerges is that every time there is a penetration into the red, the depth of the penetration, expressed by the number of item units below the red level, is recorded. If within the time frame of the replenishment time the summation of all the recorded penetrations is equal to or greater than the size of the red level, then a recommendation to increase the buffer is given. In other words, if during the span of the replenishment period the penetration of red equals the entire size of the red buffer, it is time to increase the buffer size.

Once the target level has been increased, the specific item will definitely be in the red. The increase in the target causes a new replenishment production order to be released. It certainly will take time before the new buffer size will be stabilized. Before that, there is no sense in deciding to increase the buffer again. The point here is to refrain from hasty decisions until the impact of the previous increase has been noted.11 Thus, the algorithm calls for a "cooling period" where no re-evaluation of the penetration into the red would be done. The natural time for the cooling period is one replenishment time. Therefore, it takes one replenishment time to possibly discover the buffer should be increased and another period of replenishment time until such a check should start again.

Discussion: Issues with DBM and by How Much to Increase/Decrease the Targets

The first topic in this discussion has been by how much to increase/decrease the buffer. From this question, some additional questions might be raised, such as what are the immediate ramifications for such a change, and due to them when should such changes be avoided? Shouldn't the increase of the buffer be subject to a forecast, which predicts how much the demand would grow?

In practice, the sales of one item at a specific location are too chaotic to truly support a good prediction of the quantity. However, the trend of the sales can be predicted, so we should know whether we need to increase or decrease the buffer, and decide rather arbitrarily about the size of the change.

We discuss here the behavior of sales from the manufacturer's viewpoint. In other words, wild fluctuations are less common at the manufacturer's level than at a specific store. The question is whether we'd have a better answer for the manufacturer than the arbitrary guideline that says whenever a clear signal is noted that the buffer is not adequate, change the buffer by 33 percent or any other fixed ratio that seems appropriate.

Note that the BM signal is impacted by the combination of demand and supply. When the demand goes up, the idle capacity decreases and the replenishment time gets longer. Do we know how that is going to affect the right size of the buffer?

This author's inclination is to accept the premise of having an arbitrary number for buffer increase or decrease. However, for the shop floor a decision to increase the buffer by 33 percent looks to this author like it creates too many waves in the general flow. A buffer increase of 20 percent and a buffer decrease of 15 percent look more appropriate for the shop floor. The demand from a manufacturer usually has much less fluctuation than the sales of a store, and thus the changes in the buffers could be smaller and still be able to match the trends.

Another question is what are the appropriate conditions for increasing a buffer? When the buffer is increased, the whole amount of the increase is released to the shop floor as one production order. This relatively large production order comes on top of the regular replenishments that are following the actual demand. If the current load on the CCR is high, then the last action we should take is to release another large order to the floor.

Actually, when there is high load on the CCR, it could easily generate a recommendation to increase the buffers of several items. If production management takes the recommendation to increase the buffers, then a substantial additional load will be added to an already high-load pressure. This extra load might cause more items to penetrate the red zone for too long, causing even more items to penetrate the red for too long. It can easily turn into a vicious cycle! Thus, the point is to allow buffer increases only when there is no immediate load pressure, or when additional capacity can be used for it.

Of course, buffer decreases are easier to make and they reduce the load pressure. However, if the recommendation to decrease the buffer is not justified, then some time later we would see a recommendation to increase the buffer; then, depending on the total capacity situation it might be difficult to do.

As we have seen, certain supply problems can cause "false" recommendations to increase the buffers. By "false" we mean either that the problem observed by buffer management is just a rare statistical fluctuation or that while there is a real problem with the current specific target levels, the targets should not be increased at that time. This is the case when a temporary lack of a specific raw material is causing the end item to go into the red without being able to replenish it very soon. In such a case, an increase of the buffer would not help anything. Once the missing material shows up, the dilemma of whether to increase the buffer or not can be dealt with. The main point behind the dilemma is whether the manufacturer may want to protect itself from a future lack of materials (because it happened this time) by building a relatively high level of stock. The more sensible alternative is to increase the buffer of the specific raw material and settle the issue that way.

DBM is vital for getting the right signals about the validity of the buffers. In manufacturing, the author highly recommends that any actual decision about changing a buffer should be judged by the human mind and in no way be left for a software package to dictate. Asking and understanding why a buffer is continually in the red or green should be done prior to increasing or decreasing the buffer target. In a distribution organization, the number of buffers makes such a human judgment on any buffer change very difficult.

The power of DBM comes from judging the combination of demand and supply. However, the author thinks that for the sensitive decision of increasing or decreasing a finished-goods stock buffer in manufacturing environments, a focused analysis of both the demand and the flow in the production shop floor, pointing out to possible critical changes in their respective behavior, and checking possible shortages of materials should be key in the decision. Such an analysis should be done fast, based on focused information that should be part of the information system, so that the decisions can be made quickly. Right now, such an analysis is not a part of the known TOC solution for MTA.

The Role of Protective Capacity and the Usefulness of Maintaining a Capacity Buffer

The need for protective capacity to maintain availability has already been mentioned in this chapter. A special problem of MTA is that the commitment to the market cannot be conditioned by a total amount. You certainly can tell your customers that the commitment is to maintain availability up to a certain level of one-time demand. This way you protect yourself from excessive one-time demands. For an item whose target level is 100 units, a demand of 30 units coming from one customer at one time is already problematic, and a one-time demand for 60 units cannot always be answered even when the MTA procedures are done according to the book. Therefore, it would be wise to tell customers that the commitment for availability for that item is limited to up to 15 units per customer at one time. The idea is that one-time demand of up to one-half of a zone, one-sixth of the whole target level, is still acceptable. Customers that wish to draw at one time a larger quantity should issue an order to be supplied at a certain quoted lead time, like any regular MTO type of orders.

Nevertheless, such a limit on the commitment is not capable of addressing a 20 percent rise in the total demand for all the products at the same time. After all, it is not the responsibility of any single customer to look at the total demand of all customers.

When we dealt with MTO, we established a way to deal with too much demand by being able to quote longer lead times than the standard lead time. This method of quotation smoothes the load and allows good utilization of the CCR.

In MTA, we do not have a way to restrain the demand according to capacity. Does it mean we should have to maintain enough capacity at all times? Well, it is certainly possible to sustain a peak of load for a limited time because having enough inventory on hand smoothes the impact of a temporary lack of capacity. The effectiveness of BM priorities dedicates the limited capacity to those products that need it most. However, such a peak period cannot continue for too long without affecting availability. Thus, the biggest risk to the good performance of the TOC MTA methodology is growing market demand, which requires more capacity than the CCR is capable of handling.

The planned load is a worthy tool to judge the required capacity based on the demand at hand. However, in order to judge the capacity at hand we must include in the planned load all replenishment orders. As you may recall, the regular definition of planned load for MTA takes into account only the replenishment orders that are released into the floor. Ideally, all the replenishment requests have been released to the floor, but when capacity is temporarily limited, some replenishment orders are delayed because they have lower priority and the CCR would be busy anyway processing orders that are more pressing.

Therefore, in order to monitor the overall capacity status one must run a somewhat different planned load; let's call it the full planned load that includes all production orders and also replenishment orders that are not released yet.

What we get then is the time it takes for a new production order, just released, until it is processed by the CCR. Ideally, we need it to be not more than 80 percent of the formal replenishment time. The other 20 percent represents the time required after the CCR to complete the order. When the planned load is longer than 80 percent of the replenishment time, it means the actual replenishment time is longer and thus there might be a threat on the availability if that situation continues. Therefore, if this is just a peak for a short period of time, then the system has a good chance to stay stable. However, if the market demand continues to grow, then the threat would become a reality.

If there is a reasonable assessment that the demand is truly going up, then the conclusion should be to increase the capacity as soon as possible.12 Of course, we mean to increase the capacity on the CCR, but any elevation of the CCR should initiate an analysis of whether another resource would become a true CCR, so we might consider increasing the capacity of that resource as well.

Increasing capacity is an investment, so we had better be confident that the sales are truly growing. In addition, we need to know which resources should be elevated. We know the CCR requires more capacity, but many times we have less good information on the other resources. As we'll see later, certain feedback from the floor might help us to pinpoint the resources that will need extra capacity once the demand goes up. However, additional study of other resources that might require more capacity may be needed.

There might be another way. Suppose there is a certain amount of capacity that can be purchased quickly at will; for instance, calling for overtime or even adding an extra shift. There are cases where the shop floor already works 24 hours a day, 7 days a week. Even at that rate, sometimes night shifts are not fully manned. Another way of purchasing capacity at will is by outsourcing. What is typical of all these cases is that the extra capacity costs additional money every time it is used. Moreover, to preserve that amount of capacity one needs to use it from time to time, otherwise it won't be available when the need arises. Suppose that for a whole year no extra shift was called; how easy will it be to organize it? It is not certain there is enough labor to staff it and if it is possible on paper to gather the required people, do they really wish to work an extra shift?

This author defines capacity buffer as a quick means to purchase additional capacity that is truly available on reasonable notice. It is a buffer to protect the ability of the company to commit availability and truly meet the commitment. As a true buffer, the level of use of the buffer signals the level of pressure the system is under. The use of the capacity buffer should be initiated by two different parameters: the full planned load (when it is larger than it should be) and the number of red orders relative to the average number of production orders. The full planned load approximates the real load and it depends on the accuracy of the data. However, when the full planned load is growing beyond the previous limits, one must deduce that higher than usual red orders will follow. Thus, additional capacity can be planned based on the early warning of the planned load, or wait for the emergence of red orders and then add the required capacity. Following the increase in red orders has the advantage of knowing which resources are required for expediting. This is very valuable information for a decision on investments in capacity.

The capacity buffer behaves like every buffer. The use of it could easily conform to the green-yellow-red mark. However, the capacity buffer should be in the green most of the time-meaning less than its potential is actually used. When the regular use is already in the yellow, then its function as a buffer is already compromised to a certain degree.

The Process of Ongoing Improvement (POOGI)

The fourth concept of flow13 (Goldratt, 2008) speaks about establishing a focused process of balancing flow. It certainly fits well with the way BM functions-giving the right priorities to what should be done now. However, balancing the flow should be taken seriously for the longer time frame as well. In other words, we must have a focused mechanism to identify specific areas where an improvement would really improve the overall flow.

Again, BM supplies certain basic information. In MTO every time an order penetrates into the red zone the user should enter a "reason" from a table, so monthly analysis can be done to pinpoint the most frequent reason and see what can be done to eliminate it. On top of the list of reasons, it is possible to collect the whereabouts of the production order when it turned red. The assumption is that in most cases for a resource that causes long delays there will be many times when production orders will turn red while waiting for that resource.

In MTA, entering the red has three possible causes. One is too long of a delay in order release (lack of materials or too high-load pressure). Two is too slow of a flow in the shop floor that caused the on-hand stock to penetrate the red level. Three is high sales within the last day or two. When we look to balance the flow, the third cause is not relevant. Only causes for relatively long delays are relevant for such a process.

The new suggestion by Dr. Goldratt (Strategy and Tactic tree MTS to MTA, 2008, entity 5.113.2) is to register any delay that is "too long" in a work center. The suggested definition for "too long delay" is one-tenth of the formal replenishment time. Registering such a delay does not ensure entering into the final Pareto list for picking the highest one area and trying to improve it. The other condition is that the delayed order would eventually enter into the red. Only then does that occurrence enter the Pareto list.

The feedback process requires reporting when an order arrives at any work center in the list of "to be watched" and reporting when that order has been completed at that work center. In order to implement this process, the organization needs both the appropriate software for reporting and the discipline of the operators to report it accurately.

Generic Issues in MTA

MTA for Components

The main essence of MTA is the combination of marketing and operations. Let's now consider the possible value of producing certain common components to availability. Here we are speaking of common components used in MTA finished goods for customers. Like in the original meaning of MTA, we better have a strong commitment to maintain excellent availability of those common components. The point is that in planning the production of the end products, either to order or to availability, it would be possible to rely on the availability of the common components. The value, when it is truly applicable, is a major cut in the production lead time.

The main reason for maintaining stock for common components is to shorten significantly the response time to the market. Another reason has to do with minimizing the significant setup times that are sometimes involved in producing basic parts/materials used in many end items. For instance, in many V-plants, such as plastics or paper, there are a few base materials, which are used in many end items. The primary operation in preparing the base materials (like an extruder) often has very large setup times. The operational problem stemming from the long setups is that many urgent orders for a certain base material emerge all the time, as any demand for one of the huge number of end products creates a demand for one of the base material products. The longer is the wait time for the primary operations, the more end-items and the more production orders enter the red zone, thus putting pressure on the primary operation. Urgent requests (usually for a small quantity of one base material) make it difficult to keep the minimum batch size required to keep that primary resource from becoming a bottleneck. The straightforward solution is to produce those base materials to availability, and then the only urgent replenishment is when the stock of one of the base materials is penetrating the red.

Producing base materials or components to availability splits the entire manufacturing floor into two different environments separated by stock buffers. Both are planned in separate runs, even though one environment feeds the other.

The planning and the BM for MTA of components is the same as for MTA of end items. Note two important points: 1. In order to provide a smooth transition of an item from order to availability, the initial stock must be in place.

2. Every item should be defined as either to order or to availability. If there is a need to produce the same item both to order and to availability, then define the two separate Stock Keeping Units (SKUs), one for orders and the other for availability. This point will be discussed later in the section on mixed environments.

Which Items Fit MTA and Which Fit MTO?

It is quite clear that not every item should be managed to availability. One factor to consider is the level of the demand fluctuations. Here is a simple graphical representation of a typical MTA item demand versus a typical MTO item demand. Let us start with a typical MTA shape in Fig. 10-2.

FIGURE 10-2 Semi-continuous behavior of sales for a typical MTA item.

Sales for an item that nicely fits managing to availability is having a spread of daily consumption that is less than the average daily sales (Coefficient of Variation less than 1). That also means that on most days there are some sales. Such a spread allows for frequent and fast replenishment and the on-hand stock would stay mainly in the yellow.

Other items might have a very sporadic demand that fits a MTO pattern. Most days there is no single sale, but there are days where the clients purchase relatively large amounts. It could look like Fig. 10-3.

In order to manage such an item in MTA, there is a need to hold a very large target level and in a substantial amount of the time the on-hand would be in green, which also means that DBM should not handle very well such an item. While the clients might wish to include such items for immediate availability any time they need it, the characteristics of the item are such that everybody else would have difficulty to make it available at all times. Thus, if a relatively short response is offered to the clients they would, most of the time, accept it.

The same is also true for managing common components to availability; the criterion depends on the shape of the consumption curve and the size of the spread.

FIGURE 10-3 Sporadic demand that should be better managed as MTO.

Chapter 11 is dedicated to the way TOC handles distribution, Amir Schragenheim offers a way to consider the return on investment (ROI) in carrying stock of an item. When applied to manufacturing, the same parameters still apply, and the spread of consumption is reflected in the target level necessary to maintain availability. This is a somewhat more detailed approach for consideration.

Vendor-Managed Inventory (VMI)

The whole point of MTA is to offer new business opportunities due to the extra value given to the clients, which competitors would find hard to imitate. A more particular opportunity is to ask a relatively large business client to take responsibility for the level of stock of the manufacturer's items, at the client site. This kind of business relationship is known as vendor-management inventory (VMI).14 VMI is not the invention of TOC. It is well known because some ultra-large organizations force it on their small- to medium-size vendors. Thus, it is justifiably a winlose kind of relationship. The vendor has to comply with whatever the big guy tells him to.

Understanding how to run MTA effectively raises the business opportunity where a vendor could offer to a client a desirable alternative that is typically forced on clients. We won't detail here the business relationship and how such a deal could be win-win for both the vendor and the client. From the logistical part, we need to make the distinction between standard items that are sold to many clients and items that are dedicated to the particular client for whom the offer is given.

Items that are sold to many clients should be managed according to MTA up to the plant warehouse, and thereafter by the distribution solution.