Let us look at the shop and the different entities operating in this environment.

We can categorize the sale of items in the shop as three different types20: 1. Cheetah items-these items are sold very fast relative to their stock level, enabling the retailer to reach high inventory turns21 (if managed correctly).

2. Regular running items-the items that do not fit the previous categories. These items generally exhibit moderate turns.

3. Elephant items-these items are slow movers; the retailer just can't get rid of them. These items are traditionally low inventory turn items.

What is bound to happen with the fast running items?

When items are cheetahs, by definition the market demand is high for them relative to the amount of inventory we keep for them. Regular running items (and new products) that turn out to be cheetahs are the ones most likely to be sold out. If one goes to a retailer and asks how many shortages he experiences, the most likely answer would be very few, maybe 2 to 3 percent. Misconception abounds here. If the situation and question instead were, we stand outside your store and ask people whether they found what they were looking for; in how many cases will we get an answer of "No" even though you are supposed to carry what they are looking for? The most probable answer regarding shortages would be 10 to 15 percent. This (subjective) finding suggests that the level of shortages experienced in shops is much higher than what the retailers think.22 If the typical shopping pattern of consumers is that of purchasing more than one item at a time, then the real impact of the shortages is 10-fold. How many times have you decided not to purchase an item because the retailer was missing one or two other key items you needed? You then put the items back (hopefully) and go to another retailer hoping they have all of the items. What is the chance, when having only 15 percent shortages, for a customer to find all 20 items he wants for a home improvement project in the shop? The answer is less than 4 percent, which equals (.85)20 as every single item has an 85 percent chance of being there but all 20 must be there at the same time for the purchase to be considered successful. These shortages affect the buying patterns of almost every customer.

A very interesting factor comes into play when analyzing those missing items: The 10 to 15 percent of items, the stockouts, are primarily the cheetahs! Hindsight being 20-20, if the retailer had known these items were cheetahs then he would have stocked a lot more. Therefore, lost sales he experiences is far more than the 10 to 15 percent that he might actually admit to! This is true especially in the fashion business. The retailers buy goods once at the beginning of the season for the whole season. Therefore, the fastest selling items (the cheetahs), which were impossible to predict a priori, once stocked out will be missing for the remainder of the season. For example, an item that sells so fast that all the inventory is consumed in two weeks in an eight-week season has lost sales of three times that which was initially purchased.

The elephant items represent the other side of the coin. These items are not sold as envisioned when the retailer bought them, otherwise he would have avoided them. The phenomenon that happens here is absurd-the retailer invests tremendous efforts to sell these elephant items and blocks his best display space with these items at the expense of the other items in the shop. This behavior, while expected from the psychological side, is counterintuitive in the business sense. Huge efforts that will be invested by the shopkeeper to sell the elephant items could have yielded much higher revenues from the cheetah items.

This phenomenon sometimes dwarfs the effect directed toward managing the shortages in the cheetah items!

Some industries have gone so far as adopting phrasings to hide the fact they are operating in a counterintuitive way in their desperation to solve these problems. They glorify the stockouts of cheetah items (according to TOC thinking, it is called lost sales) by calling them "sold out"! They then simply ignore the fact that the elephant items are bad for business by marking them "on sale" and investing huge efforts in selling them.

In a supply chain that is based on pull distribution, these negative phenomena are cut significantly. Recall that TOC BM is based on reacting to the actual market demand and adjusting the buffer sizes accordingly. If the market demand picks up (cheetah items), the stock buffer size will be increased, creating a mechanism that allows stockouts only for very limited time periods. That means lost sales due to stockouts of cheetah items are minimal. What of elephants? In the TOC distribution/replenishment solution, lower inventories of all items are kept, and the quantities are further decreased when consumption is low based on buffer penetration and dynamic buffering. Elephant items are much less of a problem as their quantities are initially low and are reduced further over time. Therefore, using pull distribution and DBM is very effective in eliminating lost sales and overstocking.

Some of the Finer Points in Managing the TOC Distribution/Replenishment Solution

This section details some of the finer points in the implementation. Usually, those finer points come at a later stage in the implementation, after replenishing frequently and activating the DBM mechanism, but nonetheless they should be mapped at the beginning of the implementation in order to understand better and construct the implementation correctly.

Managing Product Portfolios

To differentiate between cheetah items, regular running items, and elephant items, a simple criterion exists-the inventory turns.23 Our interest is the amount of sales of a specific item at a specific location (an SKU) relative to the inventory level of that SKU.

However, it is not enough to know the quantity in which items are sold, it is important to know their financial value as well. Just knowing from the items which are the cheetah items and which are the elephant items will not help much in driving any operational decision to improve profitability. It is important to know the magnitude in financial terms.

The goal of setting such criteria is obviously relevant when the shop owner needs to choose which items to stock. However, where a large number of items are available and the ability of each stock location to keep a large amount of SKUs is limited either by cash or by floor/shelf space, the decision is crucial. Just taking into account the inventory turns is not enough because some items are sold at such a low margin that even if they are cheetah items they contribute little to the bottom line. In addition, a certain item can be sold only once every year (an obvious elephant item), but the margin is so high relevant to the inventory investment that it is a great item to have. For the M/D, a measurement like that can be used to support the decision of which products he should eliminate from the supply chain offering. Is there a good measure for making this decision?

The best measurement for comparing which items to stock and not stock is to determine how much a certain SKU is worth keeping at the stock location. The Return on Investment (ROI) for each inventory item24 provides an excellent method of comparison across SKUs for the retailer. Retailers are usually limited by the amount of cash or space, so they should focus on the items that contribute the most to the bottom line.

Using TOC the question becomes, "How much Throughput (meaning margin) does one gain from this SKU over a year?"25 This question can be written as T = selling price TVC. To calculate the Investment, consider the following: The inventory kept at the stock location to cover immediate demand (the actual stock).

In-transit inventory to refill the buffer. The inventory in-transit is also an investment in order to protect from the fluctuations in demand and to cover for regular consumption.

Taking these considerations into account, the best number to represent the Investment needed to generate the T this SKU realizes is the buffer size. By multiplying the buffer size by the TVC (TVC/unit of SKU) of this SKU, the real inventory investment needed to generate the annual T of this SKU is realized. Note, one does not consider the timing of who owns the in-transit stock. You ordered it, and therefore there is an obligation to buy it. Hence, it should be part of the calculation.

Therefore, the formula is very simple. To calculate the ROI, the annual T of this SKU should be divided by the TVC per unit from this SKU multiplied by the (average) buffer size throughout the year.

The ROI measurement enables differentiating between three different groups of SKUs based on financial contribution: 1. Star items. These items represent a very high ROI for the retailer and certainly should be stocked appropriately throughout the chain to support the retailer. These are excellent candidates for placement at other retailers to see if consumers at those locations demand them as well.

2. Regular ROI items. These items are not in either category.

3. Black hole items. These items have a low or possibly negative ROI. These items are potential candidates for elimination from inventory. However, this is not conclusive, as some items (usually referred to as strategic) are necessary to have even though their margin is so low and/or the quantity sold is low, which places them in this group.

It is obvious that there is a correlation between the cheetah items and the star items, but this is in no way a 1:1 correlation, as is clearly demonstrated by the extreme cases discussed earlier and which will be further demonstrated by Fig. 11-8.

The decision of how to set the limit between the different groups is dependent on the specific environment, but the general guidelines are taking the top 10 percent based on ROI as stars and the bottom 20 percent as black holes. Of course, a check is needed whether these items have been replenished regularly and are not in a class because of bad management. This check shows that you have poorly managed the inventory of the black holes. One approach to improving the ROI is to reduce the investment significantly in that SKU while maintaining its T. An obstacle that must be addressed in these situations is the purchasing unit of measure; the amount that must be purchased at one time needs to be reduced. Some items are packaged in cases of 12, 24, or 48. In any situation, the shop must sell the first 11 before it can sell the 12th item. It is far more productive to split a case and possibly get three units of four different items. You then have four opportunities to sell a product to a customer instead of one. With the cash generated from reducing the inventory investment in a black hole SKU, invest it in another SKU. Another possibility to treat black hole items is to change the price of some of those products, making them more profitable if sold at the higher prices. Figure 11-8 shows an example of how these classifications can get different results.

Figure 11-8a lists 20 different items, each with its own selling price, TVC, volume, and the buffer size that had to be maintained in order to support the consumption. Assuming the buffer size was managed properly and replenishment was done properly, the calculation of the Inventory Turns and the ROI of each item appears in Fig. 11-8b. The elephants and cheetahs in the IT classification and the stars and black holes in the ROI classifications are marked. Notice that while Item02 is marked as positive in both classifications, none of the other items matches the same classification level in both cases. Especially note Item20 that achieves both a cheetah and a black hole classification, a contradiction that shows both classifications are different-ROI being the more logical one to use.

FIGURE 11-8 (a) The data for the calculation of item inventory turns, ROI, and their item classifications.

FIGURE 11-8 (b) The calculation of item inventory turns, ROI, and their inventory and ROI item classifications.

Rules for Setting up Initial Buffer Sizes

The first step in moving from push distribution to pull distribution is setting up the PWH and starting to build inventories to fill the initial stock buffers.

The decision of what size the initial stock buffer should be might seem to be a very complex decision as the amount of uncertainty is huge. Fear of making an error or the wrong decision and jeopardizing the whole initiative is natural.

The answer of how big the buffer size should be is quite simple. There are not enough words in the dictionary to emphasize the difference here between being precisely wrong and approximately right. It is not exceptional to find cases in which determining the initial buffer targets took more than three months! Starting with any initial guess and adjusting the buffer size according to DBM would have reached good enough buffer sizes much faster. Based on the parameters (demand rate and supply responsiveness), a generous stock buffer size can be determined (which is generally much lower than what is stocked now in the chain). Since the DBM mechanism will adjust the buffer sizes according to real consumption, the initial estimates are not that critical.

It is advisable to start with an initial guesstimate: taking the replenishment time from the source to the destination and multiplying it by the average daily consumption and by a factor (to cover statistical fluctuations). For the PWH/CWH, a fluctuation factor26 of 1.5 is appropriate. For the selling points, a factor of 2 is appropriate, as the fluctuations are larger there. The replenishment time to use should be: For a production environment (PWH) taking the current quoted production lead time for this item (after implementing TOC in the manufacturing environment, the lead time will usually be cut in half). Use this lead time and remember that DBM will automatically suggest lowering or raising the stock buffer level over time.

For a distribution environment (CWH, regional warehouse, and consumption points), transportation time plus something to account for a low weekly frequency of delivery if needed.

One must also adjust for the frequency of delivery. As the frequency increases, the buffer size is smaller; as the frequency decreases, the buffer size is larger. For example, it is obvious the buffer would be much smaller if the SKU was delivered daily versus if it was delivered every week.

Managing Seasonality in the TOC Distribution/Replenishment Model

Evidence supports the claim that the DBM is an excellent mechanism to monitor and control stock levels when changes in supply or demand are gradual27 or otherwise when the changes are unpredictable.

However, sudden large changes in either supply or demand are not handled well by the DBM mechanism or by any other known mechanism. An unanticipated, sudden, and steep increase in demand or deterioration in supply can cause shortages that will lead to lost sales and a damaged reputation. Alternatively, an unanticipated, sudden, and steep decrease in demand or improvement in supply will cause excess inventories and an undesired focus on sales efforts to move slow-running items.

While some of these changes in patterns are unpredictable (and thus unanticipated), experience shows that there are known recurring predictable, sudden, and steep patterns. These situations can be moderated by recognizing where and when to use the DBM methodology. The general guidelines are as follows: When the changes are gradual, either predictable or unpredictable, use DBM-this should be the mostly used method. Preferably, use automatic DBM in order to avoid mistakes and enhance the focus on exceptions.

When the changes are unpredictable and large, use DBM in order to point to the problem and use a manual decision process to define by how much the buffers should be changed when the change occurs.

When the changes are predictable and large, use the seasonality module. The seasonality module handles known patterns for sudden changes in consumption.

Known Patterns for Sudden Changes in Consumption

Most of the time significant changes in supply or demand are predictable. Marketing and Sales people know from experience when to expect these changes in demand and their consequences on the supply. Generally, the direction of the change is well known and a gross approximation of the size of the change is possible, enabling taking measures ahead of time to deal with the change. Typical causes of changes in patterns28 are divided into two groups: Pull Seasonality & Push Seasonality.

Inherent Simplicity defines the following patterns as Pull Seasonality, meaning the environment defines the demand pattern for the organization without the organization being able to do anything about it: Seasons in the year affecting the consumption of certain SKUs.

Holidays or events geographically affecting where certain SKUs will be consumed, more or less.

Inherent Simplicity defines the following patterns as Push Seasonality, meaning the organization, for various reasons that should be verified, takes actions that create a peak in demand in the market. These patterns include the following: Promotions-very similar to holidays in nature as they are short and create a spike in demand. They are generally followed by a low period in demand.

A known price increase-many times an organization will announce an increase in product prices becoming effective at a point in time. Customers generally stock up on the item before the price increase. They are generally followed by a low period in demand.

Financial period end seasonality-measurements of salespeople that focus on quarterly or yearly quotas usually create a kind of seasonality in which sales go up sharply before the period ends and down sharply at the start of the next period, caused by pulling orders ahead toward the end of the period. Note: this can also be created from the budget management of the clients-toward the end of a period, they try to take advantage of all unused budgets for purchasing. They are generally followed by a low period in demand.

Two Different Changes

Each of these situations has a beginning and an end, which Inherent Simplicity29 defines as Sharp Demand Changes (SDCs). In the previous descriptions, the beginning SDC is (usually) the event that causes an increase in demand and the end SDC signals the sudden end of the increased demand and a return to "normal" or below normal demand.

Resolving the Forecasting versus DBM Dilemma to Provide Excellent Consumption before, during, and after an SDC

SDCs present a problem in changing the buffer sizes. What would happen if DBM will continue to be used in an SDC? A possible problem is demonstrated in Fig. 11-9.

The dashed line in Fig. 11-9 (Inventory) represents the actual inventory at the site. The inventory is more or less stable until the season starts. After the season starts, since there is a huge surge in demand (Sharp Demand Increase in the figure), the on-hand inventory runs out completely. The DBM mechanism almost immediately suggests adjusting the buffer by 33 percent (from 9 to 13). The stock buffer inventory stays on zero for some time (any small replenishment orders that are already in processing are consumed immediately because the demand is so high), which triggers the stock buffer level to be increased by 33 percent (DBM first buffer increase in the figure). During this period, sales are potentially lost because demand is higher than supply. When the new replenishment quantity arrives, it is still not enough to support the new demand because the demand picked up by much more than 33 percent. That causes the same phenomenon to occur-the inventory runs to zero until the new replenishment quantity arrives at the site, representing another 33 percent increase-DBM second buffer increase (from 13 to 17 in the figure). By the time that quantity arrives, the demand has already gone down and the site is left with too much inventory to support the demand. The DBM mechanism identifies that condition but by that time, it can only reduce the buffer by 33 percent, leaving it much higher than it should be in order to properly support the demand. It will eventually reach the steady-state level again, but for some time you first experience stockouts followed by carrying excess inventories. Of course, this is an extreme case, but obviously must be dealt with.

FIGURE 11-9 The problem of using DBM with an SDC. ( 2007 Inherent Simplicity. All rights reserved.) It is apparent that sometimes crude forecasting must be used in order to avoid those negative effects.30

Identifying When an SDC Is Meaningful

A simple rule can be used to determine whether an SKU is exposed to certain seasonality effects. Look back on last year's consumption (and the year before, if possible). If one month's sales are more than twice the monthly average of the total sales (greater than approximately 15 percent of the whole year, say the Christmas season, for example), then this SKU should be looked at carefully to see whether the SKU is an SDC item. While DBM reacts to reality quickly, using seasonality forecasting does not (the shop must adjust orders manually). Therefore, it is important to define an SKU as seasonal only if it creates a huge difference with which DBM cannot cope. Most changes, especially when the replenishment time is relatively short, can be easily dealt with by DBM. If the order frequency is over a day or two and the spike in demand is high and short, you should adjust the orders manually. If ordering daily with a short replenishment time, a change as high as a 50 percent increase in consumption in the course of one single replenishment time is something with which DBM can usually cope.

Handling of an SDC

For known SDCs, you need to forewarn the upstream links in enough lead time to respond. If the spike in demand timing is known (e.g., a home football game) and big (much larger than average demand), then you should give the upstream links notification to be able to respond and plan how the SDC should be handled. When an SDC is identified, it should be treated in the following manner,31 depending on the direction of the SDC.

For a known large SDC that marks an increase in demand (also defined as Sudden Demand Increase): 1. Stock buildup.

2. Disable the DBM (cooling period).

3. Back to normal (or sometimes even below normal).