30 May/June 2010 | APICS magazine

You’ve been on your lean jour-ney for at least a year. You’ve focused on the fi ve Ss; imple-

mented production cells; and whittled down your setups by 50, 75, maybe even 80 percent. Now is the time to fi gure out how to manage your raw and outside-processed inventory as they come in via countless suppliers.

At this point, most of you probably fi nd yourselves with too many part numbers to manage and quantities that are all across the board. You manage all stockkeeping units (SKUs) using a single method, and you trust the forecast —buy what it says you need and attempt to predict any spikes or anomalies in customer order patterns. All the while, you negotiate the best purchase price from suppliers through volume. Sound familiar?

Th ere must be a better way. Perhaps you could set up kanban for all part numbers or vendor-managed inventory. Maybe you could negotiate consign-ment agreements with your suppliers. But, before beginning to work toward either of these alternatives, it’s necessary to consider demand segmentation.

Demand segmentation is the cat-egorizing of diff erent demand types into groups that share similar charac-teristics. Th is graphical representation shows the sales or consumption vol-ume of products versus their demand variability. Its premise is that all parts or products are treated uniquely. Each one has its own personal demand characteristic. When setting up buy or reorder strategies, products with high volume and low demand variability are treated diff erently than products

with low volume and high demand variability.

Traditional product quantity—also described as ABC analysis—fails to rec-ognize that high volumes are not always predictable and that low volumes can be. In trying to classify certain products for certain pull techniques, it is neces-sary to understand demand variation.

Demand segmentation analysisTo build a demand segmentation analysis of product types, you fi rst need historical sales data, preferably in 52-week buckets organized by SKU. More is better here, especially if you’ve had an increase or decrease in the past or if you experience cyclical seasonality. Th ese data may be diffi cult to come by. Consult your information technology department, or contact your soft ware

See the Relevant TraitsBetter inventory management with intelligent demand segmentation

By Shawn Kaul

MJ10_APICS_Mag_Body.indd 30 4/23/10 1:41:26 PM

APICS magazine | May/June 2010 31

provider to assist you. It’s crucial that you only look at pieces or units at this point, not dollars. You will examine inventory cost as a measure of success after imple-menting your new inventory levels.

Now refer to your historical sales data to calculate the average and standard deviation of a historical 52 weeks for every SKU. Use the aver-age and standard deviation results to calculate the coefficient of variation (CV) for each SKU. The CV helps you see the variability of your inventory order patterns. The lower the CV, the more consistent the order pattern. It’s predictable.

For example, look at an item with a low CV, such as a common screw or fastener used in multiple-assembly builds. You probably use close to the same amount of the part each month, so it might make sense to have a vendor manage this item for you. Or maybe you can negotiate a bulk stocking program with your supplier. Conversely, an item with a high CV indicates there is little-to-no predict-ability in its order pattern, so it might make more sense to purchase make-to-order (MTO) when needed.

Next, plot your data in a scatter chart arranged by SKU. Put the total histori-cal volume on the Y axis and the CV on the X axis. (See Figure 1.)

The purpose of looking at order pat-terns this way is to give you some pre-dictable data to develop a strategy for purchasing inventoried SKUs. You now have something other than a forecast to use when making strategic decisions. The CV gives your company a risk factor to use when buying component inventory.

A CV less than 1.0 lends itself to flow and pull techniques. One less than 0.5 often can be handled with rate-based replenishment methods. If a CV is greater than 1.0, it might be considered assemble-to-order. A very high CV would represent those products that you may begin to phase out of your offerings altogether—or at least convey to customers that these are custom items requiring greater lead times, which will give you ample time to order components from suppliers rather than holding these slow movers in inventory.

The bottom line is that you do not want to hold inventory of any SKU with a high CV. There’s a lot of risk in keep-ing these items in stock because you can’t predict when they will sell. Sales teams often have issues with supporting this philosophy. After all, the best way to serve customers is to have what they want on the shelves, right? But remem-ber: If you sell more of the high-CV item, it eventually will have a lower CV—and, at that point, it will qualify for a revised stocking strategy.

The next piece of information most professionals overlook when looking at demand segmentation is days of inven-tory on hand (DIOH). If DIOH data are difficult to retrieve, you still can move forward with your analysis, but it’s really helpful to chart DIOH along with historical usage and CV as a way to compare how well your organiza-tion is performing in buy strategies of slow-moving SKUs. Your DIOH data also will confirm for any doubters that using only one buying method for all product types regardless of CV is not the best method.

Figure 1 shows DIOH data plotted on the secondary Y axis. Some SKUs with a CV greater than 5.0 have an excessive amount of inventory days; whereas, SKUs with a low CV might not have enough inventories to meet demand requirements.

Now, draw several diagonal lines across the chart to help identify four distinct categories. This step can help your team come to a consensus on a strategy to use going forward. Figure 2 shows the four distinct categories.

Rate-based components typically are common in assemblies and include items such as fasteners or hardware. Because they are high volume and low CV, they are predictable in their use. Also, these items often are inexpensive and easy to buy—perfect for a vendor-managed inventory program, which frees up your purchasing team to deal with more troublesome day-to-day issues.

Kanban pull components usually are common items, but have a lower historical volume or a higher CV. These SKUs are candidates for a kanban pull strategy and can be ordered only when they meet a trigger or reorder point.

MTO items require two data char-acteristics to be considered. Some SKU quantities simply are too low to realistically have a stocking program. And some quantities exhibit too much variability to hold. The best strategy could be MTO. But remember: This is only a guideline in establishing a risk factor. There is flexibility here.

Custom, long-lead-time components often bring about the worst ordering habits and can really hurt inventory turns. At this point, it is a good idea to incorporate inventory dollars back into your data in order to spotlight the need for a change in the company’s paradigms.

What are some options for these custom SKUs? You could write off some of them, but be sure to consult your finance department for guidelines. You also can work with your sales profes-sionals to explore options such as offering sales liquidation promotions or reduced-freight programs. These can create a sales spike in assemblies and

Figure 1: Demand segmentation

MJ10_APICS_Mag_Body.indd 31 4/23/10 1:41:27 PM

32 May/June 2010 | APICS magazine

ultimately reduce some excessive, slow-moving inventory. Regardless of what you decide here, it will take some time. Stick to your plan, and be patient.

Now that you’ve determined your demand categories using historical data, your team can begin to build a strategy and policies to support its hypothesis. You also can dive down one more level to look at how to use the data to calcu-late safety stock for kanban pull items.

Calculating safety stock levelsMany professionals look at demand fluctuation and assume there’s not enough consistency to predict future demand with any confidence. They see no other way to set safety stock levels than by using trial and error—or maybe just throwing out a comfortable percentage.

Using the data in Figures 1 and 2, it would be best to use a stocking strategy for these rate-based and kanban SKUs. If your goal is to reduce inventory while maintaining or improving service levels, you must rely on historical demand variability to predict shortfalls and thus set your safety stock level according to data, not opinion.

Consider establishing a service-level factor based on the standard deviation of your historical data. Figure 3 is a positive Z-score chart commonly used in statistics to measure the distance in standard deviations of a sample from the mean. In addition, it may be used to determine service-level factor.

For instance, if you set a service-level goal of 99.9 percent, the Z-score chart shows this as a 3.09 service-level factor. Multiply that by the historical standard

deviation of each SKU to calculate the safety stock quantity required to maintain a 99.9 percent service level. It’s that easy. Using a calculation based on normal cus-tomer demand variability, it’s possible to achieve a 99.9 percent probability of not being out of stock—if you maintain the calculated safety stock quantity on hand.

When using a service-level factor and standard deviation to calculate safety stock, remember the following. First, this formula is not recommended for

all SKUs and should never be used for mid- to high-CV part numbers.

The higher the CV, the greater the variability—meaning, more inventory will have to be calculated to maintain a higher service level. For example, an SKU with a CV of 1.5 (relatively high volume) would be a good candidate for a kanban pull sys-tem. Yet, with a service-level factor of 99.9 percent, the recommended safety stock quantity may be much greater than what realistically can be held in inventory. Don’t be afraid to relax your service-level factor on an SKU to 97 or 95 percent—or until you reach an acceptable balance between the service level and inventory on hand.

The timing to recalculate your CV greatly depends on your business volatil-ity. Companies that experience high cycli-cal demand or have had recent increases or decreases in their business should recalculate CV a minimum of twice a year. You can use more than 52 weeks of data to capture such volatility, as well. Either way, this is not a formula to calculate once and then forget about. It must be managed like everything else, but it’s not necessary

Figure 2: The four categories of demand segmentation

Figure 3: Positive Z-score chart

Positive Z-scoresZ 0 .01 .02 .03 .04 .05 .06 .07 .08 .090 .5 .504 .508 .512 .516 .5199 .5239 .5279 .5319 .5359.1 .5398 .5438 .5478 .5517 .5557 .5596 .5636 .5675 .5714 .5753.2 .5793 .5832 .5871 .591 .5948 .5987 .6026 .6064 .6103 .6141.3 .6179 .6217 .6255 .6293 .6331 .6368 .6406 ..6443 .648 .6517.4 .6554 .6591 .6628 .6664 .67 .6736 .6772 .6808 .6844 .6879.5 .6915 .695 .6985 .7019 .7054 .7088 .7123 .7157 .719 .7224.6 .7257 .7291 .7324 .7357 .7389 .7422 .7454 .7486 .7517 .7549.7 .758 .7611 .7642 .7673 .7704 .7734 .7764 .7794 .7823 .7852.8 .7881 .791 .7939 .7967 .7995 .8023 .8051 .8078 .8106 .8133.9 .8159 .8186 .8212 .8238 .8264 .8289 .8315 .834 .8365 .83891 .8413 .8438 .8461 .8485 .8508 .8531 .8554 .8577 .8599 .86211.1 .8643 .8665 .8686 .8708 .8729 .8749 .877 .879 .881 .8831.2 .8849 .8869 .8888 .8907 .8925 .8944 .8962 .898 .8997 .90151.3 .9032 .9049 .9066 .9082 .9099 .9115 .9131 .9147 .9162 .91771.4 .9192 .9207 .9222 .9236 .9251 .9265 .9279 .9292 .9306 .93191.5 .9332 .9345 .9357 .937 .9382 .9394 .9406 .9418 .9429 .94411.6 .9452 .9463 .9474 .9484 .9495 .9505 .9515 .9525 .9535 .95451.7 .9554 .9564 .9573 .9582 .9591 .9599 .9608 .9616 .9625 .96331.8 .9641 .9649 .9656 .9664 .9671 .9678 .9686 .9693 .9699 .97061.9 .9713 .9719 .9726 .9732 .9738 .9744 .975 .9756 .9761 .97672 .9772 .9778 .9783 .9788 .9793 .9798 .9803 .9808 .9812 .98172.1 .9821 .9826 .983 .9834 .9838 .9842 .9846 .985 .9854 .98572.2 .9861 .9864 .9868 .9871 .9875 .9878 .9881 .9884 .9887 .9892.3 .9893 .9896 .9898 .9901 .9904 .9906 .9909 .9911 .9913 .99162.4 .9918 .992 .9922 .9925 .9927 .9929 .9931 .9932 .9934 .99362.5 .9938 .994 .9941 .9943 .9945 .9946 .9948 .9949 .9951 .99522.6 .9953 .9955 .9956 .9957 .9959 .996 .9961 .9962 .9963 .99642.7 .9965 .9966 .9967 .9968 .9969 .997 .9971 .9972 .9973 .99742.8 .9974 .9975 .9976 .9977 .9977 .9978 .9979 .9979 .998 .99812.9 .9981 .9982 .9982 .9983 .9984 .9984 .9985 .9985 .9986 .99863 .9987 .9987 .9987 .9988 .9988 .9989 .9989 .9989 .999 .999

MJ10_APICS_Mag_Body.indd 32 4/23/10 1:41:28 PM

APICS magazine | May/June 2010 33

to look at day to day, as you may do with current purchasing strategies.

A real-life exampleJoe Wilson, director of operation plan-ning for Eldorado Stone in San Marcos, California, is responsible for the plan-ning and inventory logistics of six plants and two distribution warehouses. His company historically had approached inventory management challenges from an analytical standpoint, using standard elementary statistics to estimate stock-ing targets. Wilson turned to a consul-tant in order to validate his company’s inventory management strategy.

“Demand segmentation allowed us to stratify our product offering in a more thoughtful way and zero in on the true variability in our sales,” Wilson says. “Our purchasing and inventory manag-ers [gained] insight on the importance of having our inventory levels tied to actual demand history.”

He adds that the benefits of calculat-ing Eldorado Stone’s safety stock using

statistics included sales team members being better able to meet the needs of customers and stocking the right product on shelves. With a few minor tweaks, employees constructed an inventory strategy that enabled them to reduce inventories an additional 60 percent from previous levels.

Seeing resultsIn the short term, your inventory management team will begin to realize the benefits of using historical demand data to help bring about positive and permanent solutions in inventory management. They will feel that they are accomplishing meaningful results in an otherwise temporary fix and the never-ending battle of shortages and planning mishaps. Communication quickly spreads across the purchas-ing, planning, and scheduling teams, promoting a greater good for your company and customers.

The long-term benefits are endless: Obsolete and excess inventory quickly

translates into a very high cost in any supply chain. Each additional inventory turn achieved per year can free up cash for new product development, expanded marketing and sales, modernization, reengineering, expansion, acquisitions, and debt reduction. The more inventory turns achieved, the more savings gained.

A committed change in your com-pany’s inventory management strategy—one that is based on statistically sound and proven data—will enable you to improve your working capital position and service levels. Even if digging into inventory management feels like the next daunting battle, don’t hold back. Demand segmentation is a powerful lean tool that will provide you with a fresh perspective.

Shawn Kaul is president and chief

executive officer for LSI Consulting Group

LLC. He may be contacted at skaul@

lsiconsultinggroup.com.

To comment on this article, send a message

to feedback@apics.org.

RegisteR online at apics.oRg/extRa.

APICS extraAPICS Extra Live: intelligent inventory Management

Attend APICS Extra Live to gain deeper insight into the May/June APICS magazine article by Shawn Kaul, which demonstrates how to improve inventory management practices through intelligent demand segmentation.

In this APICS Extra Live, you will deepen your understanding of demand segmentation and learn how to achieve results by defining demand segmentation categories that match your inventory to your business model. In addition, you will gain insights into managing abnormal conditions and building confidence with data.

Presented by: shawn Kaul Date: June 3, 2010Time: 1:00 p.m.-2:00 p.m. ct

MJ10_APICS_Mag_Body.indd 33 4/23/10 1:41:28 PM