A Guide to Implementing the Theory of Constraints (TOC)
Replenishment – Adding Value Through The Supply Chain
Replenishment is the method by which we add substantial value to the supply chain. We achieve this by increasing the throughput generated from the final customer – our constraint. In fact we must subordinate our whole supply chain to the constraint. That is we make sure that we have exactly the right stock, in the right place, whenever someone wants to buy it. Sound difficult? It’s not really. Not, if we think about what we are setting out to achieve.
Let’s step back for a moment to the manufacturing cases that we have examined and the step forward into supply chain. Previously, in section on production, we discussed increasing the throughput in make-to-order environments, an environment where timeliness is an explicit concern. We then examined make-to-stock and make-to-replenish environments, places where timeliness is less obvious but still a critical element. We represented our make-to-stock environment something like this;
The model consists of a manufacturing or production process where goods are transformed in some way and value is added. A make-to-stock production process is, however, decoupled somewhat from actual customer demand by a buffer of stock. Usually this indicates that the customers’ willingness to wait is less than our lead time to supply from pure make-to-order; so we must make product in advance and store it for essentially instantaneous availability. Timeliness doesn’t seem so critical, but it is. People want goods instantly.
Of course not all businesses contain a production or manufacturing process, that is, there may not be any transformation of goods carried out in the process. Instead, value is added by moving goods through both space (transportation) and time (storage) from a source of supply to the location of a demand. Rather than a flow from a process to a stock buffer as in manufacturing, we now have a flow from stock buffer to stock buffer. Let’s draw this.
It’s not too hard then to envisage a whole series of individual stock buffers feeding from one to another – nothing less than a supply chain. Let’s draw this.
Now we have four storage spaces, let’s call them nodes – it sounds far more impressive than shelf or pallet or regional warehouse – although each of these are indeed nodes on their own scale. Each node in the supply chain supplies the next node until the end-user, the final customer, makes a purchase. Also, each node in the supply chain is supplied by the previous node until the point of entry into the system, usually a production process in secondary industries or a raw material in primary/extractive industries. In effect we now have at each node, the storage space, these storage spaces are central to the replenishment solution; they are our stock buffers.
Each node could represent the quantity of a single type of stock unit in the chain as it passes from distributor to wholesaler to retailer for instance. Equally we could consider it to be the whole sum of all the different types of stock units at each of these stages. The diagram is generic; we can decide as the situation demands.
Clearly we could also have hybrid systems with both manufacturing and supply chain components, and the supply chain component might not be linear; but let’s leave these interesting facets until the distribution and marshalling pages. Replenishment is the motor for supply chain, it is the mechanism that ensures that we never miss a sale by not having the right material in the right place at the right time for the customer. Let’s have a closer look at the drivers of replenishment, but first, if you have a manufacturing background you might consider a small diversion.
Replenishment can be broadly described, or defined, as the frequent and rapid replacement of recent actual demand. The key is that there is no forecasting into the future, only the rapid response to the very recent past. We address timeliness in supply chain by the replenishment characteristics of our stock buffers. Through replenishment we can supply more goods in total, to the right place, at the right time, and most often with considerably less total stock in the system. The objective of course is to increase Throughput.
Each and every stock type at each and every node becomes its own replenishment buffer. The size and serviceability of the stock buffers is driven by how frequently and how rapidly we choose to replenish consumed stock and this is how we configure the solution. The configuration will depend upon the assumptions that we are willing to challenge about batching; both batching in time and batching of quantity.
Let’s draw this as a simple model.
Buffer management is how we monitor the motor’s performance. At a global level the stock buffers provide us with longer term feedback into the configuration when a certain degree of buffer violation or near violation becomes too common (or too uncommon) suggesting that a particular stock buffer needs to be resized to be fully effective and/or that the re-order/re-supply frequency or duration needs to be addressed.
Locally, the buffers, once in operation, signal replenishment quantities – our local prioritization system – they also absorb many small variations as well as providing day-to-day exception reporting indicating that there may be a potential stock supply violation – our local control.
Let’s add these local features to our model.
Buffer management is crucial; it filters important signals from the day-to-day noise of the system alerting us to potential problems before they become real problems, and it provides self-diagnosis that neither too much and nor too little protection is made available for each stock held.
Let’s now examine a little more about what we mean by replenishment.
Replenishment is one of those words, like for instance; quality or strategy, which means different things to different people depending upon their work environment and their experience. This is sufficient to cause quite a bit of confusion. In fact, it is probably fair to say that manufacturers have one view of replenishment and that supply chains have another view.
In general it seems that replenishment of a stock buffer is composed of two critical components;
Re-Ordering and Re-Supply
However, I will argue that there are at least 4 components and that the two additional ones should be insignificant, or at least rendered so. Nevertheless we should recognize their existence;
Re-Checking, Re-Ordering, Re-Supply, and Re-Stocking
Clearly, then, re-ordering and re-supply are the two aspects that may require the longest durations and hence impinge most upon our timeliness and the determination of the size of our buffers.
We need to examine re-ordering and re-supply separately before combining them together once again. In order to do this let’s exclude re-supply for the moment by something similar to that which applied mathematician do; “let us assume” re-supply is near instantaneous! If we do this then we can isolate some of the assumptions about re-ordering.
Let’s consider 2 re-ordering environments.
1. Fixed re-order quantity variable re-order frequency – batch lot manufacturing
2. Variable re-order quantity fixed re-order frequency – shipment lot supply chain
In fact there is a 3rd which we touched upon in manufacturing make-to-stock on the drum-buffer-rope page and which we will mention once again after considering the mechanism for determining buffer status. However, of the two above, the first, fixed-quantity, is very common in manufacturing, the world of batch lots. The second is fixed-frequency and is very common in supply chain, the world of shipment lots – and the subject of this page.
Clearly there is tremendous potential for people to say “I understand what you mean by replenishment,” when indeed the understanding is locked into the first case. I know, “I’ve been there, done that.”
So let’s work through both of these cases under the assumption of near-instant re-supply. Then we will have a look at the effect of non-instant re-supply, and how to accommodate this as part of replenishment.
If we think about it, all make-to-stock is replenishment of sorts. If we make 4000 new things that are standard items, and they don’t age, then it really doesn’t matter if we make a year’s supply and put them in the warehouse – although it would be better to be privately owned to embark on such a mission these days. Our only risk is that the demand might be for more than this plus our safety stock before we get around to producing these things again; therefore we might miss sales. If we sell less than 4000 in a year, then we just won’t make the remainder next year and bring our stock back up – right?
So in effect we did replenish the stock, it’s just that the cycle time is quite long. This is OK for companies with really deep pockets and really mundane things – industrial things. I’m not sure if we could find such a company still doing this today, well at least not on major items, but there certainly are stable established industrial firms who manufacture small volume items in their range once or twice a year. However, hopefully we are all in the business of making things to sell rather than making things to store.
What would happen then if we are supplying consumer goods or are making perishable items? Now if we make 4000 things it’s just quite possible that the market taste may change, or our competitors may bring out something ”new and improved” – even if it is only the label, or that some of the stock will pass it’s “use-by” date before it is sold. Now we risk not only missing sales if sales are greater than our forecast, we also have a very real risk of dead stock.
So what would happen if instead of making 4000 things once a year we made 1000 things once a quarter or heaven-forbid 333 things every month, or 83 and bit every week? Oops, round that up to 84 for safety – no make it 85. Can you see where we are heading? We are always replenishing whether with 4000 things or 85 things, but with increasingly smaller quantities and increasingly higher frequency. Moreover, because we are externally constrained and have processing capacity to spare, we can chase any localized market spurts, and we can also drop off quickly with any market downturns. In a word we have become responsive.
Let’s examine this graphically. Let’s retrieve our diagram for a hypothetical stock item from the finished goods section. A perfect saw-tooth diagram. In fact, it is too perfect for a processing environment. We all know that in any processing environment demand is never uniform; therefore the rates of drawdown are not uniform either. We need to take this into account. Let’s redraw it.
This looks more like reality. We replace goods with a new batch according to a signal triggered by the drawdown – the re-order point. When the demand rate is high or the replacement is slow we might drawdown on into our safety stocks a little. The frequency of replacement is driven by the rate of consumption. The replacement batch is of fixed-quantity. The maximum quantity is defined by the batch size policy that we chose.
We know from discussions of batch sizing throughout this site that reducing the process batch size will reduce the amount of inventory required to be held in finished goods and that there will be a equivalent increase in replacement frequency. Let’s draw this.
Our batch size in this example has been effectively halved and our replacement frequency has doubled as a consequence. We are replenishing with smaller quantities more often and maintaining the service level with half of the previous inventory. Let’s go to the next step and halve process batch size once more.
Now the batch size of each replacement is one quarter of the original batch size and the replacement frequency is 4 times the original. Clearly we are moving towards replenishment, less and less stock needs to be held on hand and it is replaced more and more frequently.
Effectively as batch size decreases the maximum stock in absolute terms approaches closer and closer to the re-order point stock level. Of course the ideal batch size would be a unit of 1 – single piece transfer/just-in-time.
Therefore replacement in a processing environment is characterized by;
Fixed Re-Order Quantity & Variable Re-Order Frequency
The process requires a finished goods stock – but it should be as small as possible to decouple the customer from the process. The customer can still get stock immediately, and the process replaces it with as small a batch as possible.
In re-order point systems the batching policy is explicit. However, in some min-max systems the batching policy may be more implicit; “the batching policy is hidden under the size of the max minus the min (1).” The batch size is defined by the physical and especially the policy constraints of the process.
Most manufacturers will recognize the above discussion as what they term replenishment – and this type of fixed-batch replenishment does find its way into supply chain too. However, it generates two traps that we should be aware of and try to avoid.
The first trap comes from synchronization. Consider the following.
Within any two layers of a supply chain there may be a one-to-many relationship. In a min/max system we need to consider what happens when more than one downstream node hits the re-order point at about the same time? The upstream node experiences a “wave” of demand. Improbable? Not at all.
How could such a synchronization arise in the first place, from simple chance? Or did it arise from the last synchronous re-supply from the upstream node? This seems far more probable. Often times we create our own problems for ourselves. This type of problem creates large waves in upstream nodes when gentle and continuous downstream consumption is the reality.
The second trap comes from the number of layers or levels of nodes in the supply chain.
In a min/max system it can take quite some time for a signal for replenishment to move back from a single downstream node to an upstream node. If we multiply this effect by several layers or levels then the delay soon becomes quite considerable (2).
Combine these two situations, a significant number of layers and some one-to-many relationships and then even a simple linear supply chain is suddenly not so simple any longer.
Let’s look at the alternative then; fixed re-order frequency with near-instant re-supply; a more common replenishment system found in supply chains.
In supply chain, as opposed to manufacturing, it is possible to break out of fix-quantity re-orders and instead fix the re-supply frequency. So we need to investigate this as well.
Let’s start again at the beginning once again with our perfect saw-tooth graph.
In this instance we no longer have a re-order point based upon quantity. Instead we have a re-order date. Re-ordering is done at some regular interval; once a month at the end of the month, once a week at the end of the week, once a day at the end of the day. In the absence of manufacturing batching constraints the imperfection that we are seeking in this graph and which we would indeed see in the real world is that the replacement amount will be variable rather than constant as we have drawn. Let’s have a look at that.
Total inventory oscillates below some full value that we have determined as necessary to protect customer demand. The re-order dates occur at fixed-frequency; once a day or once a week or once a month or some similar measure. The difference between the full value and the current value at the re-order date determines the amount of the re-order.
Just as in the processing environment where we could reduce the quantity in each batch, here we can increase the frequency and obtain the same effect; reduced inventory of finished goods stock without degrading service levels. Let’s halve the re-order date duration and thereby double the re-order frequency. This is the equivalent of ordered once a fortnight instead of once a month, or ordering once a week instead of once a fortnight.
Now we have half the inventory, and twice the rate of replenishment. The replenishment frequency is fixed, but the replenishment amount remains variable. Let’s show the effect of doubling the frequency once more.
Now inventory levels are one quarter of the initial value and yet we can maintain our service level by replenishing 4 times more often than in the initial case. This corresponds to a reduction from monthly orders to weekly orders or weekly orders to daily orders.
Therefore replenishment in supply chain environments is characterized by;
Variable Re-Order Quantity & Fixed Re-Order Frequency
Many times the variable quantity that we re-order to in fixed-frequency re-ordering is in actuality a forecast value for future expected consumption. Many of the ERP systems that generate these targets are operated on a monthly cycle ensuring large quantities that are produced infrequently.
The two traps that we mentioned for min/max systems still operate to some extent. In terms of the one-to-many relationship and synchronization, if all the nodes re-order on the same date, this will cause a “”wave” of demand on the upstream supply nodes when the customer drawdown might have been quite gentle and uniform during the preceding period. In terms of the number of layers of levels of nodes, it can still take a loooonng time for a signal from the customer node to arrive back at the source node . Depending upon the order dates it could take as many replenishment cycles as there are nodes for a point of sale signal to arrive back at the first node.
None of this will surprise anyone in supply chain. But what can we do about it?
We know how to significantly reduce the size of a stock buffer by either smaller quantities or more frequent re-supply. Classically we achieve one by fixing the other. Hopefully this analysis has allowed manufacturers to break the fixed batch quantity mind-set that is sometimes carried over into supply chain. In supply chain we fix the frequency and this is the path that we will follow here in our development of replenishment buffers.
Let’s have look, then, at this thing called a replenishment buffer.
Let’s examine buffers in a different fashion to the saw tooth diagrams that we have used to date. Let’s draw some fixed-frequency replenishment buffers over a complete cycle and see if that further helps to draw the distinction between the two environments of operation – processing and supply chain. Let’s first look at normal demand.
We start at the beginning of a cycle with a replenishment order having just been received. Over time we have drawdown of stock until at some pre-selected and recurring date we place a re-order, and some time later that order is in-turn received and stock returns to its replenished state. This is our stock buffer. We have divided the buffer into 3 zones. Zone 1, the red zone, represents the material that manufacturers would call safety stock. We expect stock levels to oscillate between zone 3, the green zone, and zone 2, the orange zone.
Let’s have a look at this system if we started with a relatively less stock in the buffer to being with but at the same rate of drawdown.
Because we start with less stock in the buffer we replenish with a larger amount and it returns to the same level as in the first case.
What then if we were to start with a relatively more stock in the buffer to begin with but at the same rate of drawdown? Let’s look at that.
Now, we find that the difference between the buffer at the re-order date and the full buffer is less and thus the replenishment amount is smaller. However, the buffer is still replenished to about the same level as in the first case.
In this particular example there appears to be some robustness about the buffer design with respect to the rate of stock drawdown. Therefore let’s look at both a substantially greater and substantially reduced rate of drawdown.
What happens then during a period of heavy drawdown? Let’s use our original case but we will increase the rate of drawdown by 50%. Let’s have a look at that.
During a period of heavy demand we might see the buffer stock move into zone 1, the red zone. In this case we would need to make efforts to ensure that the replenishment will arrive in good time – we might expedite existing re-supply orders to some extent. In fact crossing into the red should raise a flag that demand is heavy at the moment. We find that our replenishment takes us to a position that is lower than we started with, and the buffer will stabilize here over additional cycles.
Let’s then have a look at periods of light demand. Again we will use the initial case but we will decrease the demand by 50%.
We can see that during periods of light demand the buffer is only slightly depleted. The finished goods buffer may remain in zone 3 or pass into zone 2 a little. We don’t need to do anything here. The replenishment takes us to a position that is higher than we started with, and the buffer will stabilize here over additional cycles.
On the drum-buffer-rope page I tried to make a conceptual distinction between fix-frequency buffers of make-to-replenish and the classical fix-quantity buffers of re-order point and min/max of make-to-stock. I suggested that fixed-quantity min/max buffers might be viewed as filling from the bottom and that fix-frequency replenishment buffers – like we have just drawn above – might be view as filling from the top. If we fill from the top, then we need to know just how big the buffer is. We need to know how to size the buffer. Let’s pool our knowledge of decreasing batch size and increasing shipment lot frequency and address how to size a supply chain replenishment buffer.
A supply chain replenishment buffer must protect the downstream node, or the customer demand if it is the last node, from fluctuations in both downstream demand and in upstream supply during the period of interest. The period of interest is the replenishment time. The replenishment time we already know is composed of the re-order duration and the re-supply duration.
However, so far we have fairly much assumed near-instantaneous re-supply. Let’s continue to do that for a while because I think that it will help us to understand better the various components of replenishment. Let’s start with a simple case – two nodes.
Let’s assume that we order once a month and that the stock comes from across town the next day – that is pretty much instant re-supply. How large should we size our stock buffer then? Here is a formula (3).
Buffer Size = Average Recent Demand In
The Re-Order Period
An initial suggestion for a good value of safety is plus 50% (3). Such a suggestion automatically creates a buffer zonation of 2/3rds to 1/3rd (more on that later).
A more recent verbalization for the buffer size is; the maximum forecasted consumption within the average replenishment time, factored by the level of unreliability of re-supply (4). It seems that the term “re-supply” was however used loosely to mean all of the replenishment duration rather than a specific component of replenishment duration as we are using it here. This was corrected in the Insights program on distribution and supply chain as; the maximum forecasted consumption within the average replenishment time, factored by the level of unreliability of replenishment (5).”
Thus the more variable the customer demand, the bigger the buffer. The more variable the re-order duration, the bigger the buffer. The more variable the re-supply duration, the bigger the buffer. The higher the required customer service level, the bigger the buffer. These situations are better accommodated in the later verbalizations. With this in mind, however, let’s stick with the first formula; it will help us to better understand the situation if we limit ourselves to this more straightforward case.
Let’s assume some numbers. Let’s assume that we sell 1,000 units per month on average. With near-instant re-supply our buffer size is 1,500 units. Lets draw that.
What, then, if we order weekly under the same conditions? Our average will be around 250 units per week, our safety is 50% so we now stock 375 units. Let’s draw this.
Basically we reduced our buffer by a quarter by re-ordering weekly and receiving it next day from across town. Of course, the variability around the average for weekly demand will be greater, but then the time between re-supply during which we may dip into our safety is much reduced. So we shouldn’t get lost in statistical theory when pragmatism will suffice.
What if we can’t get re-supply next day from across town? Well, then we have to make the buffers bigger. Let’s have a look at that.
What if we order monthly but the shipment comes across country or across the sea? We have to take that into account too. Let’s start with monthly re-ordering and a month to re-supply.
Our initial assumption was for a 1000 unit average demand per month, then we must have 1000 units to cover the monthly re-order period, plus we also have 1000 units in transit, and to this sum we must add our 50% margin. Near the end of the month, just before we start to re-order, last months order will arrive. Sound familiar? So, looking at the buffer sizing decision we now have this;
By the end of the month our buffer looks somewhat like this.
But what if we are unhappy with the absolute size of the resulting buffer? Then, we must either re-order more frequently and/or re-supply more quickly. Let’s look at the impact of more frequent re-ordering with the same supply duration. Let’s re-order weekly for our monthly supply duration.
The buffer size declines to 1875 units and at the end of a week it looks somewhat like this;
Reality Check. Although we have instigated weekly re-orders, from a static perspective we may be mislead to believe that for an 250 unit average consumption per week we must maintain 625 units as a margin. The dynamic reality is that over a month we still sell on average 1000 units with 625 units as a margin – compared with 1000 units as the margin beforehand. Clearly, more frequent re-ordering has some value. However, we should strive to re-supply more quickly, the impact is much greater. Let’s do one more iteration to show this. Continuing with weekly re-orders and quicker, fortnightly, re-supply. We have the following.
Now we have 500 units of goods-in-transit and a stock buffer size of 1125 units. At the end of a week our buffer looks somewhat like this;
From a dynamic perspective, the sale of 1000 units of per month is protected by a margin of 375 units compared to the 1000 units of margin in the first instance. If we could go to the next step of weekly re-order and weekly re-supply (a condition that we examined in the previous section) we would be looking at a buffer size of 750 units – a quarter of the initial case of monthly re-ordering with monthly re-supply. The margin that we need to maintain would reduce to 250 units, also a quarter of our initial condition.
But let’s not forget, the objective is not inventory reduction per se, it is getting the right stuff to the right place at the right time so that our throughput goes up, so that our customers come back to us for more (and not our competitors), and so that our throughput goes up even further.
It seems any time someone makes a suggestion for smaller batches and more frequent delivery, someone else will raise the issue that operating costs will go up. There is a very good discussion by Cole on why operating costs should, in fact, go down while addressing this particular aspect and a myriad of other aspects as well (2, 3). However, please also check out the “truck and trailer” analogy in the production page on batch issues. There are graphs there that are relevant to supply chain. From these we can infer that it is very likely that a significant proportion of the active inventory in any supply chain is in actuality the smaller part of the total shipment lots. More frequent re-supply of this portion of the stock will have considerable leverage upon the whole system without substantially increasing the total shipment lots.
Once we have determined the buffer size, how do we determine the replenishment amount?
The only pieces of information that we have are the buffer quantity that we are aiming for, the quantity present as stock-on-hand at the re-order date, and any goods-in-transit. Thus the replenishment amount is the difference, and as depletion will continue to occur until replenishment is achieved, then the replenishment when it does arrive will be to a less than the full buffer.
Re-Order Quantity = Buffer Quantity – Stock-On-Hand – Goods-In-Transit + Orders
The re-order quantity is fairly much the same as the proportion that we have been calling consumption above. In fact it should be fairly clear that the re-order quantity replenishes consumption.
Of course it would be easier to write “re-order quantity = consumption” but the amount of consumption only has relevance with respect to the buffer quantity that we establish. So it has to be determined by difference.
But where did that “orders” term suddenly come from? Well, “that would be the computer!” You see often times the order processing people put an order against a stock unit which remains currently unpicked but nevertheless committed to a customer. The most obvious cause is waiting for transportation. We could leave the term out, but seeing that we have made a firm commitment to sell it in the near future we may as well replace it sooner rather than later.
If your experience is confined to supply chain, it may seem unusual to even consider why the replenishment buffer should be determined by anything other than quantity. However, on the drum-buffer-rope page, we labored the point that in a manufacturing process the size and activity of the constraint, control point, assembly, and shipping buffers are measured in units of time. This is a unique feature (#). It is a consequence of the acknowledgement of the existence of a singular constraint within a process. For constraint buffer activity we said that;
At a manufacturing constraint an hour is an hour but the number of units may differ.
The number of units differs due to the fact that different types of product using the same constraint may use different amounts of constraint time. The unique perspective brought about by the recognition of a singular constraint allows us to define the length of the buffer in time also. Essentially the buffer is sized and “sees” the duration from the gating operation to the constraint due date. Moreover the buffer “sees” committed demand – work that has been released to the system.
Why then, doesn’t this carry over into supply chain replenishment if it is so important in drum-buffer-rope manufacturing? Well, in a way, it does as we shall see. Replenishment buffer size and activity is defined by time but is measured by quantity. Why is this?
Let’s examine replenishment buffer activity first.
The clue comes from the buffer sizing discussion earlier when we used the “average recent demand.” Even when the re-order and re-supply periods have military regularity to them, the actual amount will vary because the demand rate varies over time. Therefore, the amount of time or demand that our material stock can “cover” is variable. Sometimes demand is high and we utilize units faster and over a shorter period. Sometimes demand in low and we utilize units slower and over a longer period. But a unit is still a unit. Thus;
At a replenishment node a unit is a unit but the hours may differ.
In replenishment where nothing is undergoing conversion or combination – there is no processing – then we use units of quantity to measure buffer activity.
What about the replenishment buffer size then?
The unique perspective brought about by the designation of a replenishment node allows us to define the length of the replenishment buffer in time. Essentially the buffer is sized and “sees” a duration that extends through one period of the re-order and re-supply cycle. However, the buffer now “sees” uncommitted demand – we can not tell how much we will sell in the next hour or day or whatever. Therefore, we must once again substitute “non-variable” units of stock in place of the “variable” amount of time or demand that they “cover.” Thus the replenishment buffer size is also measured in units of quantity.
Both manufacturing and supply chain buffers are defined by time; the period that the buffer “sees.” However, in supply chain we measure buffer size and activity in units of material, and not in units of time.
We called replenishment the motor for supply chain solutions. Re-order and re-supply duration determine the buffer size and are the characteristics which configure the system. Buffer management is the monitoring and control function that we use once the configuration is in place.
Schragenheim describes the role of buffers and buffer management in replenishment as follows. ”Instead of trying to be very precise in very uncertain situations (like using sophisticated forecasting techniques), TOC strives to build a robust design that is good enough. The initial parameters are based on crude forecasting that is complemented by crude assessment of the variability of both market demand and replenishment time. What complements good-enough planning is a very flexible and priority-driven execution control system that is capable of taking care of the exceptions.” “Buffer management is a control mechanism. The idea behind it can be summarized as: identifying situations where the planned protection is almost exhausted (1).”
Buffer management is a benign form of control, basically, it is reporting by exception. Thus we focus on the few things that are locally important and we don’t focus on many things that are locally unimportant.
There are two mechanisms by which we can manage our buffers. The first and older method is buffer zonation, Schragenheim and Dettmer call this “traditional buffer management (6).” The second and more recent approach is buffer status (1). Schragenheim and Dettmer described the precursor to buffer status (in a manufacturing make-to-stock environment) as the “S-DBR approach” to controlling uncertainty and variation (7).” Let’s work through the buffer zonation approach first, leading onto a discussion on local performance measurements, and then we can examine the more recent buffer status approach.
In fact we have mentioned buffer zones in passing throughout this page. As with the manufacturing logistical solution (or project management for that matter), the buffer is divided into three. These are referred to – from nearly full to nearly empty – as; zone 3 (the green zone), zone 2 (the orange zone) and zone 1 (the red zone). It refers only to the stock-on-hand, the stock that we have that we can use to immediately fulfill a new order.
Let’s show this.
Most often buffers should oscillate between nearly fill, zone 3, the green zone, and partially empty, zone 2, the orange zone. The exception reporting occurs when there is a penetration into the red zone of the buffer, that is, when our stock-on-hand drops to less than the red zone quantity. Let’s look at this.
A zone 1 stock buffer penetration offers us two sets of information. The first is a simple record of an incidence – we have a hole.
The response to this should be to ensure that more material – goods-in-transit – are not too far away and can be restocked in time before the buffer is fully depleted.
The second piece of information is the magnitude of the depletion. Now things become more interesting. Unlike a make-to-order manufacturing environment, where each job is discrete with an absolute delivery date, and the magnitude of a buffer hole can be quantified when the job is finally completed, make-to-stock is continuous – we continuously replenish and customers continuously deplete our replenishment. So, we can not put a mark in the ground that says “hole closed on this day.” Instead we must continuously monitor the number of units missing from zone 1 and the number of days that they are missing for.
We must record the number of units “missing” from zone 1 each day that any units are missing and we must aggregate this over our reporting period – be that a day, a week, or a month. The measure is unit-days late. Note that it is an absolute measure based upon the quantity of material and the amount of time. Now we must put a value to this measure. Let’s look at local performance measures.
There are two local performance measures. They are measures of local unit-days-late, which we have just examined, and local unit-days-wait. We can attach values to these in for-profit systems and call them throughput-dollar-days late and inventory-dollar-days wait.
We know the fundamental or system-wide measures for a for-profit system; they are throughput (sales minus totally variable costs excluding direct labor), operating expense (including direct labor), and inventory (all capital investment). However we can’t and don’t want to apply these to sub-systems or else we are going backwards to a reductionist/local optima approach. In a supply chain however, it may be that each subsystem is indeed a separate business with its own throughput, inventory, and operating expense. Nevertheless we can still use the two local performance measures to ensure alignment of the whole supply chain.
If we can’t supply an order ex-stock at the time requested then that order is considered to be late. We operate the system to take account of dependencies and variation so we would hope that nothing is late. The measure of lateness is the value of the goods (throughput) multiplied by the number of days late to obtain throughput-dollar-days. It seems that we can apply this measure to two different places. We can apply throughput-dollar-days lateness value to zone 1 buffer penetration for internal measurements (8) and order non-fulfillment for external measures (4). For the external measure we should aim for a zero value. Thus, whenever there is a hand-off within the supply chain we can expect to use a value of unit-days-late.
Let’s draw this situation for a simple supply chain.
The other danger that we must guard against is that we might inadvertently build stock (a more polite term for channel stuffing). We can guard against this with a measure of unit-days-wait. In a for-profit situation we attach the value of the goods and multiply by the number of days resident in the system or the subsystem to obtain inventory-dollar-days. Increasing the amount of inventory or causing the same amount of inventory to sit for longer will result in the measure going up. We should aim for a constant or declining value for this measure for a given level of output. Let’s add this to our simple diagram.
Measures of system and subsystem waiting-in-process and lateness are a simple way to ensure alignment with the system’s goal. You will find that you are measuring these situations already, but that with these units it will make more sense.
Schragenheim uses the term “buffer status” to refer to buffer consumption or buffer penetration (1); where buffer status is:
Buffer Status = (Buffer Quantity – Stock-On-Hand) / Buffer Quantity
In this situation the buffer status value is the inverse of the proportion of stock-on-hand. Let’s draw this.
This presents the buffer as a proportional continuum rather than as absolute quantities in 3 discrete zonations. Importantly; different buffers of different sizes can be compared directly, one against another. Let’s show this.
The zone 1 level, sometimes called the emergency level (1) or red line level (7) represents an exception that should trigger an action. When on-hand stock is less than the zone 1 level efforts must be made to expedite completion of the re-supply of that item as soon as possible. The default level for zone 1 is one third of the replenishment level (1). Thus a buffer consumption value of more that 66% indicates an incursion into zone 1.
Consider the buffer status of 75% for stock item B above. In absolute terms this represents an incursion of (0.75-0.66) by 100 = 9 units.
Consider the buffer status of 70% for stock item D above. In absolute terms this represents an incursion of (0.70-0.66) by 500 = 20 units.
We might be tempted to pay more attention to the goods-in-transit for stock D than the goods-in-transit for stock B, but stock B is more important, the buffer status is worse.
Cole considers that there are 3 reasons that might lead to a resizing of a buffer once it has been set (2). They are;
(1) Buffer is too small
(2) Buffer is too large
(3) Seasonality or promotions
Using Schragenheim’s criteria, a buffer is too small if penetration into zone 1 is too frequent and too large if penetration into zone 1 is too rare (1).
Using average values of consumption and re-order and re-supply as we have here, it might be assumed that half the time our zone 1 would be too small and half the time it would be too large. In fact this is correct, but the period over which it concerns us is near the end of each re-order period. So, for instance, if we have a monthly re-order period we may penetrate into the red zone – near the end of the period. But how long do we do this for? If it is one or two days out of 20, then we have a zone 1 penetration of 5-10% of the total time – not 50% of the time as we might first expect.
Let’s look then at seasonality.
There is a demand for electric fence energizers for pastoral farming. The electric fences – often a single wire set a little below a person’s hip height – keep cattle in the pasture eating what they should eat and out of the areas that they shouldn’t eat. And although cattle can stretch their necks a surprisingly long way, on the whole this is a very effective system. Dogs who insist on holding their tails upright while passing under these fences will also attest to their efficacy.
However, grass grows at the greatest rate in spring, and this is often when electric fences are used to “break-feed” new pasture. Thus the majority of the demand for electric fence energizers occurs within 2-3 months. How do we ensure that we sell all the energizers in all the model configurations that are desired in the quantities required without either leaving some demand unsatisfied (because then the sale may fall to a competitor) and without leaving excess and unsold stock to carry for at least the next 9-10 months?
We need to set a seasonal buffer for this example. Firstly why don’t we estimate the maximum expected sales that we could hope to make by the end of the peak sales period. Secondly we need to determine the available production capacity during the peak period. We then subtract the amount that can be produced during the peak sales period from the maximum sales amount and this becomes our pre-peak target buffer. We build to this buffer prior to the season and then once the season is underway we can chase the peak with our existing capacity. We can therefore be assured that we won’t get left with too much excess stock to carry and we are also assured of meeting our most favorable estimate for sales.
We examined this topic on the drum-buffer-rope page in the production section, but let’s revisit it here. Supply chains can occur “free standing” as we have examined then here, or they can occur in combination with manufacturing either after or before a manufacturing process. Now that we have a much better understanding of how to utilize replenishment buffers lets re-check this understanding against raw material or inwards goods stock buffers. Let’s draw a basic diagram of the situation.
We need to be clear that we should never allow a vendor to become a constraint. To do otherwise would mean that our ability to generate throughput is controlled by someone else.
Vendors, unfortunately, are one of those groups were it is very easy to externalize the challenges of managing the system and blame the vendors. Well, we can do that. However, ask yourself, will it cause anything to improve? We might even be able to change vendors – until the next vendor goes “bad” on us too. Vendors really are beyond our span of control and most often our sphere of influence as well, so about the very best thing we can do is to buffer ourselves against their impact upon our business. We can modify our basic replenishment formula to account for vendor behavior. Here it is.
Buffer Size = Maximum Expected Demand
In The Longest Re-Order Period
Such a more conservative buffer sizing; “maximum expected demand” and “longest re-supply period,” may cause some inventory levels to go up, but we have to ask ourselves whether we are in the business of making money through sales or in the business of saving money through reduced inwards goods. If we are still not happy, then we should try and persuade our vendor to own the goods until we consume them. That way the vendor will learn something about the cost of their own (un)reliability. Better still, we should try to reduce the inventory size (if necessary) by increasing the frequency of the re-order and decreasing the duration of the re-supply while fully protecting our own throughput.
How can we compare what we have learnt here with traditional supply chain management? Maybe it is best to think of current supply chain management as the reductionist/local optima solution for distribution, just as MRPII is the reductionist/local optima solution for processing or critical path is for project management. Traditional supply chain management uses forecasting to try and overcome the long lead times. The long lead times generate large work-in-process but somehow as a consequence we never have all the right things in the right place at the right time. Instead we have challenged the very need for such long lead times. No matter how good or how expensive our software solution is, if it is using traditional reductionist philosophy; we won’t be able to put the right protection the right place to drive system profitability.
The only way to overcome that is to use replenishment buffering, we can best evaluate the impact by examining the two traps that we discussed earlier. The first was synchronization caused by a one-to-many relationship between source and demand.
More frequent replenishment means smaller buffers, it means smaller and more frequent re-order quantities, it means that source and the various demand nodes are in communication much more frequently. The actual re-supply quantity responds to local demand, and therefore overall demand at the supply node is much more likely to be an even flow, (the average or rather the aggregate of all of the local demands) rather than an episodic surge.
What then, if we have several layers of nodes in our system?
Again, as the re-order and re-supply periods are reduced, so too is the period between a distant customer action and the time it is felt at the source node. Instead of months, it might be a matter of days. Moreover, it could be as little as a single day. Let’s have look at that.
What if as well as communicating back to the next upstream node, we also communicated back directly from the point of final sale to the source node? Then we could ensure that the whole supply chain “knows” the current customer demand as quickly as is possible. This allows the source node to be far more flexible in following actual demand while each downstream node is fully supplied.
If we are willing to challenge that we must forecast and that re-order and re-supply times must be in-frequent then there is much we can do to ensure that the supply chain functions with precision and profit.
We have looked at 2 replenishment environments;
1. Fixed re-order quantity variable re-order frequency – batch lot manufacturing
2. Variable re-order quantity fixed re-order frequency – shipment lot supply chain
But there is a third
3. Variable re-order quantity and variable re-order frequency
We touched upon this in the drum-buffer-rope page in the section on manufacturing make-to-stock and discussed the rules required to action it. Within a manufacturing system it is possible to launch new stock orders at any time based upon a new stock order release priority determined by the relative stock order status for each stock buffer. Schragenheim showed how this becomes a self-regulating system (1). There exists a potential to extend this mechanism into supply chain; especially those that are computerised.
Consider once again that replenishment is composed of the following;
Re-Checking, Re-ordering, Re-Supply, & Re-Stocking
It stands therefore that whenever re-checking is more frequent than re-ordering, and re-ordering is already frequent, that there may be re-checking periods when there is no re-order quantity required, or the re-order quantity for a re-order period is smaller than one that we wish to commit to. In this case our re-order frequency has become variable rather than fixed. The more frequently that we are able to re-check and re-order the more likely that our re-order at any one time for any one stock may be less than we wish to commit to.
Thus we have variable re-order quantity as before, but we now also have variable re-order frequency as well. This appears to me to be a natural progression as shipment lots become more frequent and smaller. It is an ideal which we should aim to work towards.
We can summarize this as a simple table.
Using correct buffer sizing criteria we can ensure that we can always supply the next node in our supply chain to the desired level of service regardless of the variability or reliability of our vendors or of our customer’s demand.
Moreover, reduction of the re-order duration and reduction of the re-supply duration can reduce the overall inventory in the system substantially while increasing overall service levels. Nevertheless, the objective is not inventory reduction per se it is increased throughput.
Next let’s see how we can apply this to two particular configurations, divergent supply chains in distribution, and convergent supply chains in marshalling.
(1) Schragenheim, E., (2002) Make-to-stock under drum-buffer-rope and buffer management methodology. APICS International Conference Proceedings, Session I-09, 5 pp.
(2) Cole, H., (1998) Implementing distribution – layers 4-5. Video JMT-16, Goldratt Institute.
(3) Cole, H., (1998) Distribution – layers 1-3. Video JMT-6, Goldratt Institute.
(4) Goldratt, E. M., (2002) Theory of Constraints self learning program of distribution and supply chain. CD-ROM, Goldratt’s Marketing Group.
(5) Goldratt, E. M., and Goldratt, A., (2003) Insights into distribution and supply chain. Goldratt Marketing Group.
(6) Schragenheim, E., and Dettmer, H. W., (2000) Manufacturing at warp speed: optimizing supply chain financial performance. The St. Lucie Press, pp 123-135.
(7) Schragenheim, E., and Dettmer, H. W., (2000) Manufacturing at warp speed: optimizing supply chain financial performance. The St. Lucie Press, pp 175-207.
(8) Schragenheim, E., (2003) Measures and trust in SCM. PowerPoint presentation link to CIRAS Center for Industrial Research and Service - Iowa State University.
This Webpage Copyright © 2003-2009 by Dr K. J. Youngman