WO2005122076A2 - Managing an inventory of service parts - Google Patents

Abstract

An inventory of service parts may be managed by assessing a company's capability, optimizing the inventory and implementing a management program based on the assessment and optimization. Optimizing target stock levels for the inventory of parts may be accomplished by calculating an inventory baseline for understanding information about the currently held inventory; developing a service strategy for a set of segments; quantifying a service level for each of the segments; analyzing the segments and their service levels for identifying at least one logistically distinct business; assigning each of the segments to a "best-fit" planning model for indicating each segment's deployment, replenishment, forecasting and review characteristics; identifying a probability distribution function for estimating a demand process of each of the segments; and calculating a target stock level for each segment.

Description

MANAGING AN INVENTORY OF SERVICE PARTS

BACKGROUND OF THE INVENTION

[0001] This invention relates to inventory systems and specifically to planning and deploying inventory systems for service parts used to service and repair equipment.

[ooo2] Many companies regard post-sale servicing of their products as little more than a distraction. From this perspective, service is only an expense. However, analysis shows that post-sales service can be a significant revenue generator when properly planned and managed. For example, in the highly competitive jet engine business, manufacturers have realized that the value of servicing a product over its life can exceed the original sales price by as much as five times.

[ooo3] The tasks faced by post-sales business groups vary greatly by industry and by customer type. Personal computer manufacturers may have a large client base but only a relatively small number of parts to keep on hand to service perhaps 30 models of PCs. In contrast, manufacturers of construction or land moving equipment may have a smaller client base but may need to service such a wide range of complicated machinery that perhaps 500,000 replacement service parts need to be available to the repair technicians. Tracking and planning for the half million parts is a very challenging task.

[ooo4] Unfortunately, those managing the inventory may not be well qualified. For example, certain OEMs allow their dealers to control part inventories. The planners at these dealerships may treat all parts equally, applying the same forecasting, stocking, lot sizing and reviewing policies— regardless of the demand, supply and profit characteristics of the individual parts. For example, planners may make no distinction between an item with a high-volume demand, stable order patterns and quick replenishment lead times, and another item that rarely fails in the field, is sourced from one supplier and has a six-month lead time. As a result, these dealers may keep excess inventory aging on shelves while lacking the specific parts needed by their service technicians.

[ooo5] In an attempt at a solution, some post-sales business units have implemented software to help their inventory management needs. However, current transactional and advanced planning software fails to identify and integrate the detailed root causes of inventory performance. For example, demand for service parts typically exhibits random, or stochastic demand patterns; this uncertainty must be included in developing deployment and replenishment strategies. Also, enterprise resource planning systems currently available may help OEMs to meet only 40% - 60% of their high-volume post- sales needs. This leaves a 60% - 40% gap that has been difficult to fix.

[ooo6] What is needed is a way to close the gap between the inventory of service parts and the need for the parts. What is needed is a method to prioritize the management of a parts inventory to optimize the process, thereby having the proper mix of inventory to meet agreed upon service levels without overstocking on inventory. What is also needed is a way to perform such management without requiring planners to have an advanced degree in statistics or operations research. What is needed is a methodology that is cost-effective as well as scientifically based rather than only being based on rules of thumb or 'guestimates'. What is needed is a way to determine the drivers of inventory levels for service parts and to control these drivers. In addition, what is needed is a systematic and easy-to-understand methodology and toolset for that will allow the average user to conduct advanced optimization techniques on their service parts inventories.

[ooo7] There are several other factors and issues typically associated with managing service parts inventories. First, supplier performance for service parts is typically very poor; replenishment lead times typically range from 6-18 months, and line fill rates typically are less than 50%. Next, service parts distribution networks are typically fragmented with lots of individual locations: central depots, field depots, customer depots, and mobile stock. The large number of forward deployed inventories makes inventory visibility difficult, thereby making it very difficult to develop and implement stable deployment and replenishment strategies. Next, many service parts are also considered repairables, which are parts that can be fixed when they fail, instead of disposed. Repair operations require reverse flows (from customer to repair depots), forecasting of repairs, and disposition decisions, all which further add complexity to managing service parts.

BRIEF SUMMARY OF THE INVENTION

[0008] In one aspect of the present invention, there is a method for optimizing target stock levels for the inventory of parts, such as those used by asset operators and equipment manufacturers to service their equipment. Under this method, a service strategy may be developed for a set of segments and a service level may be quantified for each segment. The segments and their service levels may be analyzed for identifying at least one logistically distinct business. Each of the segments may be assigned to a "best-fit" planning model for indicating each segment's deployment, replenishment, forecasting and review characteristics. A probability distribution function may be identified for estimating a demand process of each of the segments. In one embodiment, identifying the probability distribution function may include calculating a target stock level for each of the segments. Such target stock levels may be calculated to likely meeting a service level at a desired cost objective.

BRIEF DESCRIPTION OF THE DRAWINGS

[ooo9] Figure 1 is a flowchart of one embodiment of the invention divided into five steps.

[ooιo] Figure 2 illustrates how sampling may represent an ABCD population.

toon] Figure 3 is a flowchart of the one embodiment of the invention's sampling methodology.

[ooi2] Figure 4 is an illustration of pipelines identifying a logistically distinct business. [0013] Figure 5 is a diagram of a planning model continuum.

[ooi4] Figure 6 shows one approach to calculating the distribution function as contemplated by the present invention.

[ooi5] Figure 7 demonstrates how distribution functions may offer insight.

[ooi6] Figure 8 illustrates a structured analysis for the stocking decision.

DETAILED DESCRIPTION OF THE INVENTION

[ooi7] 1. Calculating the Baseline Inventory

[ooi8] Figure 1 shows a flowchart of a general steps in one embodiment of the invention, which includes calculating the current baseline inventory, developing a strategy and a series of segments, assigning each segment to a planning model, matching the demand for service parts with a distribution function and then calculating optimized target stock levels.

[ooi9] The present invention's process may begin by calculating an inventory baseline 110 for understanding information about the currently held inventory. While gathering this baseline is known in the art and there are various techniques that can be used, some of representative tasks are to gather initial inventory data, build the inventory baseline model and validate, modify/customize data requests to operating environment, determine data sampling strategy, identify data sources within the IT infrastructure, and submit detailed data requests.

[0020] A goal of the inventory baseline is to understand the present inventory: such as by answering what is on-hand, where is it, what are its characteristics and how well is it currently operating. While it is not technically difficult to generate the baseline, it can be a difficult task to handle. In many industries, service parts are not tracked once they are distributed to the repair technicians. In such a case, the inventory of service parts may be considered an operating expense. Thus, once perhaps 30 to 50 percent of the inventory is sent ahead to satellite depots or to individual technicians, that inventory becomes invisible and outside of the equation.

[0021] Since tracking down the inventory piece by piece may be unduly challenging, software may be used to create statistically valid samples of the inventory to approximate the inventory baseline. This can create a representation of the inventory while minimizing collection time and effort. In some environments, one may extract transactional data from each network echelon and location to model the entire supply chain. To generate the sample, a planner may determine the total parts population to be sampled and then determine if a sampling strategy is needed (since sometimes analyzing the entire parts population may not be too difficult). If a sample is needed, then the planner may create a statistically valid sample from the population and identify randomized stock keeping units ("SKUs") to be in the sample. Then data for those SKUs may be extracted from various data sources and the statistical analysis performed.

[0022] Figures 2 and 3 illustrate one novel way to create a proper sample. This approach reflects design decisions made about the present invention's optimization approach. Figure 2 shows a parts population as a pie chart 205 that can be represented by sample 210. In figure 2, the parts population is segmented into four classes A through D. Such a classification may be referred to as ABCD. In such a classification, the parts assigned to class A make up 80% of cumulative dollar volume for a company 215, class B has parts representing the parts such that class A and B make up 95% of cumulative dollar volume 220. Similarly, class C includes the parts such that classes A, B and C make up 100% of cumulative dollar value 225. Class D may include those parts that make up 0% dollar volume 230.

[0023] One skilled in the art will recognize that the parts population may be divided in various ways without departing from the scope of the present invention. For example, some in the art commonly divide populations into three classes, known as ABC. One might also divide the classes based on a criteria other than collar volume. Thus various numbers of classes and various class criteria may be chosen. Regardless of how the classifications are chosen, a sample 210 may statistically represent the profile of the parts in the population 205.

[0024] One may use various methods to generate the sample. Figure 3 illustrates one approach that may be used in one embodiment of the present invention. While figure 3 assumes that the population and sample have four segments, namely A, B, C and D, one skilled in the art will recognize that the method may be used for other segmentations.

[0025] At step 310, the total population of the parts to be included [N] may be determined. The scope of the analysis may be defined at this point, such as the number of locations or echelons, the number of internal vs. external sites, active vs. inactive parts, etc. At this point, one may choose to distinguish SKUs from raw product numbers.

[0026] At step 320, whether or not a sampling strategy is needed is determined. This decision may be driven by the analysis tool to be used. For example, Excel may only process approximately 65,000 records while a database tool may process many more than that number. If the population [N] is small enough, sampling may not be needed.

[0027] If a sampling strategy is not needed (step 324), then the entire parts population [N] may be modeled or analyzed without the use of a sample population. Otherwise, the process of figure 3 may continue in order to generate a statistically valid and randomized sample (326). At step 330, ABCD unit volume segmentation may be conducted on the entire parts population to determine the population count for each category, NA, NB, NC and N_D. Such segmentation may have already been done. Or, if necessary, the category populations may be estimated by applying percentages to the overall [N].

[0028] At step 340, the sampling error, or level of precision to be used in the analysis [e] may be determined. In some circumstances, the sampling error chosen may be 90% (.10) or 95% (.05). Of course, other sampling errors may be used. In the analysis, [e] may be the probability measure that states how much the sample characteristics )such as the mean and standard deviation, for example) may deviate from the population if [N] had been analyzed instead of [n].

[0029] At step 350, the sample sizes for each of the segments are calculated. For example, ABCD segments would have (ΠA, ΠB, nc and no). One way to calculate these sample sizes is with the formula:

[0031] where e may be 0.05, 0.10 or another preferred value and where x represents a segment, such as A, B, C, or D as discussed above. Thus, if four segments are used, then the formula is applied four times to derive a sample size for each of the segments. One skilled in the art will be aware that this formula is a simplified version of more advanced sampling techniques and that other formulas may also be used.

[0032] At step 360, the product numbers in each segment are sorted or ranked based on some chosen criterion measure. For example, unit cost and item name are two possible criteria. Once the product numbers are sorted, then each part number may be assigned its ranking number. For example, the first product number in segment A may be assigned to 1. If this is not feasible to system limitations or otherwise, then steps 360, 370 and 380 may be skipped.

[0033] At step 370, a random number generator is used to randomly select the parts to include in the sample from each segment. For example, if Excel is being used as the sampling tool, then the formula "=RANDBETWEEN(1 , 333)" may be used to generate the random numbers (where the current segment has 333 elements, for example). If using Excel, one may wish to use the Excel Analysis Toolpak add-in. The quantity of random numbers needed to be generate matches the values of ΠA, ΠB, nc and no.

[0034] At step 380, the randomly selected parts are identified and their data is extracted. For example, if NA JS 2,000 and ΠA JS 333, then one may generate 333 random numbers between 1 and 2,000. Then one may match those random numbers against the sorted, numbered A items. These may then become the sampled parts for that category. [0035] As another way to minimize the effort of calculating the baseline, the present invention offers a statistical analysis tool with data requests preconfigured for different environments. For example, data requests for wholesale distribution, retail aftermarket, airline MRO, telecomm maintenance, high-tech spares, dealer channels, public utility/energy, military logistics, fixed asset maintenance and plant operations may all be installed. The invention contains data requests that collect the data in a method that facilitates the optimization approach in the invention.

[0036] At the end of the baseline inventory procedure, some of the deliverables may be the inventory baseline model, completed data request templates, and a data sampling and management strategy.

[0037] 2. Developing a Strategy and Defining Segments

[0038] Developing a service strategy 120 for a plurality of segments is another step of the present invention. While it is shown in figure 1 after step 110, actually it may be done before, during or after that step. Broadly speaking, this step involves viewing the parts inventory from a top-down view and understand a strategy for what kinds of service levels the company wants to offer to its customers. One of goals of this step is to select a segmentation criteria that is based on business requirements and that may create meaningful and unique segments that can be analyzed for operational insights.

[0039] Two examples of segmentation criteria are a supply-focused segmentation and a demand-focused segmentation. In a supply-focused segmentation approach, the chosen segments may be based on criteria such as: supplier lead time variation and duration, supplier delivery performance, standard cost of part, replenishment frequency, economic lot size, supplier relationship type, and part lifecycle phase. In contrast, in a demand-focused segmentation, the segments may be based on: criticality of customer demand, degree of demand perishability, part profitability or contribution margin, economic importance of customer, strategic importance of customer, demand variability, and interrelation with other segment businesses. [0040] Results from segmentation may be used to find "pipelines" within the customer's business to understand how to manage a segment in a more focused manner. Figure 4 shows one example of pipelines and how logistically distinct businesses ("LDBs") may be identified within the broader parts supply chain. In a complex supply chain that is tightly managed, there may only be a single LDB. Other supply chains may indicate the presence of more than one LDB. In figure 4, a "churn and burn" LDB is indicated.

[0041] LDBs are critical to structure the customer service strategy of a service business for three reasons. First, construction of LDBs allows the business to identify the unique physical flows that exist in any distribution network. For example, the LDB modeling may demonstrate that a specific grouping of customers also exhibit a specific ordering profile, such that they are driving the majority of the logistics activity in the network. Second, LDB's also allow the business to identify the unique service and supply chain requirements associated with that LDB. For example, LDB modeling will identify the lead time, fill rate, packaging, delivery, etc. requirements that are associated with a logical grouping of customers. Last, LDBs are important because they determine the value component associated with a logical grouping of customers and parts. For example, LDB modeling may show that a group of customers drive the majority of sales, are the least profitable accounts, have the highest service requirements, but are also strategically important to the business and therefore require new selling and operating capabilities to manage.

[0042] Segmentation may also be described from different viewpoints. For example, high-level customer segmentation models may be built. Part segmentation models may able be constructed. For other companies, it may make sense to identify the service levels by channel, customer grouping, SKU, etc.

[0043] Once the segments are identified, a planner may use empirical data to quantify a service level for each of the segments. Assigning the service level attempts to balance the cost of holding or carrying the inventory for the segment against the cost of a stock- out in which a needed service part must be ordered. Service levels may be defined implicitly or explicitly. With the implicit method, one uses a calculation to imply the optimum service level as a function of stock-out costs and carrying costs. Such an implicit service level may be a best-fit if the values for the input parameters to the function are available, if the business has not existing customer service strategy, or if the business does not understand cost-to-serve concepts. When the service level is implicitly determined, the planning planner considers costs and service to determine the optimum balance for a customer or LDB. Presently, most companies use the implicit method.

[0044] The explicit service level approach uses management expertise to set the acceptable minimum number of planned stock-outs. While in the past a company's sales force may have instructed the parts inventory planner to maintain perhaps a 98% fill rate, there was no tie back to the company in terms of cost. The explicit service level approaches provide such information.

[0045] Such an explicit approach may be valid if the business management is capable of assigning discrete, differentiated service levels to customers. Three such explicit levels are: cycle service level, fill rate level, and ready rate level. In the cycle service level measure, a specified probability of no stock-outs per replenishment cycle is calculated. This is generally known as an availability measure and can include the probability of periods having zero demand. This cycle service level may be a best fit with numerous periods of zero demand.

[0046] The fill rate service level is a specific fraction of demand to be satisfied routinely from the shelf. This is the most common measure and assumes no backorders or lost sales. It may be a best fit for higher-volume parts or for distribution systems where replenishment lead times are fixed or very costly to change.

[0047] The ready rate service level is a specific fraction of time during which net stock is positive. It may be the least common measure, but has application in emergency environments. It is complex to use to determine optimal inventory policy, but may be applicable for end-of-life products. [0048] One embodiment of the present invention combines these approaches and may leverage the implicit method for critical and high-value items while using one of the explicit methods for less-critical or less profitable parts.

[0049] 3. Assigning Planning Models

[0050] Referring again to figure 1 , the next step may be to assign each of the segments to a "best-fit" planning model 130, which is a decision driven by the requirements and rules that define each segmentation class. Such a planning model describes the deployment, replenishment, forecasting and review parameters for the segment. Figure 5 is a diagram of a planning model continuum that illustrates the variation in the planning models from the most basic (on the left) to the most advanced approaches (on the right).

[0051] This continuum indicates that rather than attempting to offer the highest level of service for all parts, one can be more surgical by choosing plans along the continuum that best fit each segment. For example, for low churn items, having the deployment plan be a rule of thumb (such as a time-based criteria with 10 days coverage), may be adequate since reordering can be done quickly. Thus, a replenishment strategy for such a plan may be to order every 5 days. Forecasting may be accomplished using historical sales (with a weighted moving average). The review strategy may be for the planners to review such rules of thumb once a quarter.

[0052] At the right end of the spectrum of figure 5, some segments may need a complex plan. While the cost of this type of plan may be greater, it may be justified due to the critical nature of the parts, the high profitability of the parts, etc. This type of planning model has been supported by commercially available software since around the year 2000. Some of the many vendors offering software to support such complex plans are: SAP, Servigistics, Finmatica, i2, Manugistics, MCA Solutions, Baxter Planning Systems, and Xelus.

[0053] One of the goals of this step of choosing a best-fit planning model for each segment is to focus planners on what is important rather than having them try to manage all service parts equally. Planners may now spend the proper time and effort on the proper segments for an optimized planning approach for deployment, replenishment, forecasting and review.

[0054] 4. Determining a Probability Distribution Function

[0055] Once the best-fit planning model is assigned, the fourth step in the process may be to identify a probability distribution function ("PDF") 140. In this step, one may use a range of statistical tests to identify (i.e., "fit") the demand process to the most likely probability distribution that represents the demand. This can be a difficult mathematical procedure, but generally one may collect demand data and then find the probability distribution function that underlies the part and model that function to gain insight from it.

loose] As one would expect, if demand is normally distributed (such as a bell curve), then managing the stock fora part is generally easy. Unfortunately, part demand is not always normally distributed. Prior part systems for inventory management did one of two things. The first type of prior part systems skipped this step of determining the distribution. Rather than understand the complex function describing the distribution, a planner calculated target stock levels for the service parts using rules of thumb or other rough approaches.

[0057] The second type of prior part systems used a back door approach to attempt to generate a number associated with the probability distribution function. In such prior part techniques, one used historical demand data for the services parts and used that data to simulate the stocking locations. The simulation would be run to find the first pass fill rate ("FPFR"). Once the FPFR was known, one would run the simulation again on an iterative basis, slowly (and clumsily) backing into the number hopefully associated with the distribution function. This technique was used without understanding the function itself. Such systems had many disadvantages. In addition to not revealing a probability distribution for the demand, distributions with probabilistic functions, stochastic functions or randomized functions were generally missed. By directly calculating the probability distribution function, the present invention overcomes the shortcomings of such prior art systems.

[0058] Another prime disadvantage of using this back door approach is that it is not repeatable or consistent. For example, if this trial-by-error method is used and the optimum stocking level is determined, it is most likely by random chance; the dynamic nature of service parts supply and demand will make this same result impossible to recreate in the next planning cycle. Likewise, if the back door approach is used and over- or under-service conditions results, there is no ability to determine a root cause of the event because there was no formality in determining it in the first place.

[0059] Figure 6 shows one approach to calculating the distribution function as contemplated by the present invention. First, a planner may collect demand data 610. Preferably, data is collected for a part segment or LDB to represent multiple, similar SKUs. In one embodiment, monthly demand data for one to three years may be collected. To minimize the data needs, a manageable set of candidate demand processes may be selected and used as proxies across other, similar demand processes.

[0060] Second, the data may be analyzed for insights 620. Histograms may be generated to visualize the shape and skewness of the distribution. The data may also be analyzed for auto-correlation errors and independence may be assessed.

[0061] Third, distribution fitting tests may be performed on the data 630. In this step, statistical software may be used to perform fitness tests. In one embodiment of the present invention, 22 predefined probability distributions may be compared with the fitness tests. Such tests may generate a relative score (out of 100, for example) based on the distribution parameters.

[0062] Fourth, comparative analytics may be run to select the distribution function 640. Graphical overlays between the histogram and the probability density function may be performed. Fitness tests (such as the Chi Square test, the Kolmogorov-Smirnov test, and the Anderson-Darling test) may be run against specific probability distribution functions. Based on the results of these and other tests, the final distribution may be chosen for use in the deployment algorithm. While the tests involved in figure 6 are well known in the art, such traditional planning approaches for service parts fail because they apply a standard distribution to a non-normally distributed part. The present invention overcomes this deficiency.

[0063] Using the present invention's method for step 140, a planner generates the function associated to demand and can therefore gain insight from it. For example, if the function is stochastic, the planner would realize that the part number would need to be managed more closely and rigorously. To illustrate the insights that may be gained, refer to figure 7. The top distribution function is a normal bell-curve 710. In such a distribution, one may calculate the reorder point and the safety stock 730 that should be kept on hand in inventory. However, if an inventory part is represented by a Poisson function 720, then the mean and variance the are same value, the reorder point is at a different location and the safety stock that should be kept on hand 740 is much greater.

[0064] The approach described here in step 140 may be less costly to apply than the brute force, back door approach of the prior art systems and the findings may be more accurate, which translates to cost savings for inventory management.

[0065] 5. Calculating Target Stock Levels

[0066] As a final step shown in figure 1 's general view of the invention, the target stock levels ("TSL") for the segments are calculated 150 based on the deployment algorithm. This step of the process may answer questions for deployment and replenishment. For example, for deployment, the TSLs may answer how much inventory is required to ensure an adequate level of service. For replenishment, the TSLs may answer when inventory should be order or moved, and by how much.

[0067] To calculate the TSLs at this step of the process, specific algorithm policies may be applied that may perform well over a wide range of demand values and types. Such policies are valuable because in the traditional planning approach to parts inventory management, a z-score is calculated for an area under the distribution curve. The shaded area (see figure 5 ) represents the demand that the planning meets while the unshaded portion represents the amount of inventory that will be stock-outs. Since current systems cannot calculate non-normal distributions readily or accurately, a z- score may not be accurate. The policies in this step offered by the present invention convert the probability distribution function into a planning language that is usable by planners. For example, the planner may plug in the monthly demand for a SKU in units as well as the variation in units to calculate an inventory plan for the part. The accuracy of the present invention may double the savings of inventory costs over the prior art.

[0068] Two such algorithms that may be used are (S-1 ,S) and (s,Q), where "S" stands for inventory level and the first parameter is the inventory level and the second parameter is the order quantity. In this case, (S-1 , S) is a stocking model where the system will order one unit (S) when one unit is used and falls below the level of S. (S- 1 ,S) is also called "issue one, replace one." This algorithm may perform well for low demand and sporadic demand parts, such as a satellite hub for telecomm networks and parts closets for a utility service. On the other hand, the (s,Q) policy may perform well for moderate demand and high volume demand parts where it is desirable that replenishment lot sizes may vary. For example, dealer parts inventory and plant maintenance stockrooms may make the (s,Q) appropriate. These two models explicitly account for demand variability, supply variability, and service requirements. The policies are easy to understand, easy to implement, and may be automated.

[0069] Once the TSLs are calculated using such models as (S-1 ,S) or (s,Q), a planner may also need to use a structured analysis to make a stocking decision (i.e., to stock or not to stock) for individual SKUs. Figure 8 illustrates one example of a structured analysis for the stocking decision, based on an airplane company.

[0070] The foregoing description addresses embodiments encompassing the principles of the present invention. The embodiments may be changed, modified and/or implemented using various types of arrangements. Those skilled in the art will readily recognize various modifications and changes that may be made to the invention without strictly following the exemplary embodiments and applications illustrated and described herein, and without departing from the scope of the invention, which is set forth in the following claims.

Appendix

[0071] The appendix section of this provisional patent application is a practice aid (in the form of a slide presentation) for one embodiment of the present invention.

[0072] The Appendix starts on the following page.

Key Messages

For large OEMs, asset operators and resellers of capital goods equipment, service parts inventories typically represent significant financial investments. In addition, the time and location availability of these inventories is the primary driver of customer service. Unfortunately, many service businesses have not rigorously or formally managed these inventories - in other words, taken a scientific approach. As a result, both the financial and service performance of these businesses has suffered. Accenture's offering contains the specialized, hard-to-acquire content a business needs to simultaneously improve the financial and service performance of their parts business. It's Accenture's experience that a pure technology solution is not the answer to optimizing these inventories - the right mix of skills, processes and technology are needed to transform the parts business.

Document Overview

The primary purpose of this document is to provide a market offering to our clients to drive financial and operating value in their service parts operations - Provides the toolkit to optimize their service parts planning capabilities - Presents deliverables, benefits and project approach - Presents Accenture client case studies

The secondary purpose of this document is to act as a training tool and job aid for Service Management personnel - Provides details on our methodology - Provides application-specific content for project support

This document also communicates to a broad Accenture audience the capabilities of the Service Management team in service parts management

Accenture's Approach

©

Accenture's Assets

Client Examples

Ό σ> ϋ CO

Industry requirements determine the roles service parts play - from hedging against supply and demand risks, to complying with regulatory requirements, to enabling the sale of profitable service contracts. Roles of Service Parts

The value of service parts inventories for many businesses is large, and is often a significant determinant of overall capital utilization and cash flow. Potential Free Cash Flow Impact from Turns Increase Parts Annual Investment¹ Turns² Of 30% Of 60%

- North American $525M Commercial Airline 1.1 + $121W! + $197 t

²Turns performance determined by Accenture duπng project work

Consequently, these unique requirements have forced companies to deploy unique capabilities and to acquire specific skills - all in an attempt to optimize parts inventories. Unique Capabilities ^■ ggsg Physical Network

Multi-echelon network Managing all-time buys / Demand segmentation: Dynamic rebalancing across management minimum order quantities known vs. unknown network locations ("fair play") Mixed-mode transportation (MOQs) Forecasting techniques for Solving with large number of management Strategic sourcing for parts sporadic demand item-location combinations Storage locations at Supplier relationship Incorporation of installed Item criticality considered customer sites / VMI management for parts base data Advanced algorithms for slow- Differentiated service levels Lead time compression Demand correlation analysis movers

Managing varying order Integration between OEM End-of-life planning for profiles by season, market production planning and retirement and disposal and customer service planning Super-cession planning Supplier direct shipments Collaborative dealer channel Part substitution / Import-export management forecasting and interchangeability replenishment Provisioning for new product S&OP for parts launch

Industry leaders have been able to align their service parts business with the objectives and intent of their overall corporate strategy. Representative Capabilities and Companies

Based on Accenture's experience, we have identified common, reoccurring themes on parts management across industries and clients.

Typical Issues Common Business Responses

Based on Accenture's experience, we have identified common, reoccurring themes on parts management across industries and clients, (continued)

Typical Issues Common Business Responses ■ Sales and Marketing plans created without ■ Inventory personnel have to over-provision Operations input or collaboration to avoid missing sates Miliini ll ■ Engineering mandates 100% parts availability at ■ The business builds-in obsolescence beginning of new product launch expectations in its financial projections

Sparse documentation or visibility to configuration Technicians will carry excess truck stock Configuratioi of installed equipment and the business expects extra truck rolls βffir No on-going updates to configuration as compensate for lack of configuration maintenance occurs knowledge

Status of open purchase orders and delivery dates Local personnel over-order and hoard parts

Based on Accenture's experience, a parts business will benefit from Scientific Service Parts Management® if they can answer "yes" to three or more of the questions below.

1. Is customer service dropping (fill rates_* delivery times, back order duration, number of split orders) while inventory investment is growing? 2. Does the business struggle with answering questions such as, "what should my service levels be?", "how much inventory do I need to support that service?", and "where should parts be stocked in the network?" 3. Is the business using a simplistic, single-echelon stocking network, that is Incapable of offering

© differentiated service levels to specific customers? 4. Are planners and buyers applying the same broad set of rates for deployment and replenishment for all parts, even though the SKU base is non-homogenous? 5. Does obsolete Inventory account for 5% or more of gross inventory? 6. Is there no integrated planning process between the functional departments that influence inventory performance, such as Sales, Marketing, Engineering, Logistics, and Inventory Management? 7. Has the business adopted a mindset that parts forecasting is too difficult to accomplish, and therefore has given up on a demand planning capability? 8. Does the parts business suffer f om adverse supplier actions, such as poor fill rates, lengthy delivery lead times, and mandated minimum order quantities (MOQs)? 9. Are there forward-deployed inventories that are not under perpetual control, or that the business lacks visibility to, such as technician stock or customer site inventories? 10. Are expediting and non-recoverable costs (premium freight, indirect labor, overhead) too high, from fulfilling too many emergency and split orders?

The primary objective of Scientific Service Parts Management® is to optimize the planning capabilities and processes of the parts supply chain, all in a sustainable fashion. Critical questions will be answered with hard facts and analysis.

Broad Issue Area Specific Questions Answered c What unique customer segments does the service business serve?

- How do requirements vary by segment? ^■- What service levels should I offer to different segments? : How should we manage unique customer-part combinations?

What service parts-specific capabilities does the business need?

How is the parts business currently performing? How can the business drive new revenue or cost and asset efficiencies? What is the value of improving service parts capabilities?

. How should we manage replenishment of the inventories? What metrics and targets should be used to manage parts operations?

^'1

υ iy/* $B Q.

Offering Scientific Service Parts Management^® to a client can typically take one of three

Our approach is comprehensive - it focuses on improving inventory performance along with all the related capabilities necessary to sustain the improved performance. Capability Assessment Track

Phase 1 - Phase 2,-. Phase 3'- Modeling and Analysis Pilot and Test Impjenient » "To identify the true drivers of service parts inventory levels, and to develop "To pilot the new planning "To broadly strategies to mitigate these factors in a sustained fashion." models and processes in a deploy new controlled environment." capabilities to ^B "To use statistical tools and models to optimize target stock levels based on begin value service requirements and operational characteristics." "To adjust and stabilize the new realization." models where needed."

The capability assessment can be conducted concurrently, or prior to, inventory optimization; the objective is to create the infrastructure that will sustain improved inventory performance. 1. Develop Initial Improvement % Pejrform Quantitative and 3. Refine Hypotheses gind Byϊld Hypotheses Qualitative Assessment Business Case Tasks Tasks Tasks m Determine functional and ■ Gather and process data ■ Host working sessions with key technology scope of capability ■ Conduct interviews and site visits personnel

Ul assessment Ul ■ Perform inventory analytics and ■ Continue to build and refine ■ Develop initial hypothesis tracking benchmarking hypothesis tracking document document ■ Perform qualitative best practice ■ Finalize business case ■ Construct evaluation tools - assessment ■ Finalize inventory analytics interview guides, site visits, data requests ■ Construct initial business case ■ Finalize best practice assessment structure and rationale ■ Submit data request and schedule ■ Finalize interviews and site visits site visits and interviews Deliverables Deliverables Deliverables ■ Interview write-ups/summaries ■ Working session documents ■ Assessment scope model ■ Site visit write-ups/summaries ■ Refined hypothesis sheet ■ Initial hypothesis sheet ■ Complete data request B Finalized business case ■ Data request templates, interview ■ Initial set of inventory analytics, β Finalized analytical package with guides benchmark report and best practice detailed quantitative and qualitative assessment Interview and site visit schedule assessments ■ Business case structure and format

The capability assessment can be conducted concurrently, or prior to, inventory optimization; the objective is to create the infrastructure that will sustain improved inventory performance, (continued)

Based on prior inventory optimization projects, the team will quickly develop an initial set of hypotheses regarding improvement opportunities.

Of course, as the analysis is conducted and key personnel provide their input, this starting point will be refined and validated.

The quantitative and qualitative assessment will be driven by the inventory analytics, interview guides, and best practices that are components of our toolkit.

Representative Analytics Safety Stock - Supplier Lead Time Reduction Modeling

Inventory Management ¹ Deployed inventory segmentation Parts movement profiling ^j SKU growth and proliferation ^c Network inventory fragmentation

Demand Management Forecast Accuracy vs. Safety Stock Tradeoff Analysis Forecast accuracy and bias analysis Safety stock vs. forecast accuracy tradeoff Demand variability analysis Supply Management Supplier complete order fill rate analysis

- Supplier lead time performance analysis Safety stock vs. supply variability tradeoff Demand Forecast Accuracy and Bias Analysis Fulfillment and Logistics Parts depot performance modeling ^r Order profile analysis "^• Shipment consolidation analysis

All of these tools are focused on the unique characteristics of the service parts business - not manufacturing operations or the finished goods supply chain. Representative Best Practices and Interview Guides

Process: Inventory Management Sub-process: Deployment Planning and Execution Description of Current Recommended

In addition to the qualitative assessment, our methodology will determine baseline performance and then use industry benchmarks to identify potential improvement opportunⁱtⁱes. Representative Performance Benchmark: Cost-Service Curve

5 6 7 8 10 Parts Months of Supply (Gross)

Using input from key personnel and our analysis, a final set of hypotheses are confirmed; those that cannot be validated are rejected or changed.

These validated hypotheses will feed the inventory optimization track and act as the foundation for the capabilities necessary to sustain the improved inventory performance.

The recommendations from the capability assessment are presented in a master plan format; the projects are prioritized based on management criteria and/or their support of sustainable inventory optimization.

Capability Prioritization Master Plan Shows relative importance of capabilities Shows interdependencies of capabilities Can be customized based on management's Customized based on business needs and objectives funding requirements Allows for multiple factors to be considered Good tool for aligning business investments across functional areas

The primary objective of the capability assessment is to identify the true drivers of inventory performance; oftentimes, its discovered that inventory management personnel have little

The inventory optimization track will update the value of the planning parameters used for deployment and replenishment, and will then migrate each SKU to its proper planning model. 2, Develop Service Strategy and 3. Assign to "Best-Fϊt" Planning 1. Gather Operational Data Segmentation Models Model Tasks Tasks Tasks ■ Gather initial inventory data * Format and cleanse data ■ Develop unique material management strategies for each ■ Build inventory baseline model and ■ Determine parts and customer parts classification validate classification strategy and > Deployment strategy ■ Modify/customize data requests to segmentation criteria operating environment ■ Build segmentation models and > Replenishment strategy ■ Determine data sampling strategy validate > Review strategy ■ Identify data sources within the IT ■ Build cost-service model > Forecasting strategy infrastructure ■ Develop the procedures, tools, and m Submit detailed data requests Deliverables rules to be used by planners ■ Parts and customer classification Deliverables strategy with segmentation criteria Deliverables driven by business requirements m Inventory baseline model m Completed analysis showing ■ Parts and customer segmentation ■ Completed data request templates migration of SKUs to new planning models with defined service levels models ■ Data sampling and management ■ Cost-service equation/model 3 Comprehensive package of details strategy ■ Analytical outputs and conclusions on new SKU-level planning and from segmentation exercises execution requirements

The inventory optimization track will update the value of the planning parameters used for deployment and replenishment, and will then migrate each SKU to its proper planning model. (Continued) 4. Identify Probabilistic Demand S. Calculate Target Stock Levels Processes (TSL) Tasks Tasks ■ Review new best-fit planning ■ Execute deployment algorithms for models to determine which SKUs each SKU to be modeled to are driven by stochastic demand generate TSL ■ Obtain demand/usage data for ■ Selectively validate TSL against these parts historical ownership levels ■ Apply distribution-fitting analysis to identify probability distributions Deliverables ■ Validate probability distributions ■ Validated models (either working against demand processes model or frameworks embedded ■ Identify deployment algorithms to with working algorithms) to support use for stochastic processes on-going TSL calculation ■ Selective validation of TSL outputs Deliverables (SS, ROP, ROQ) for specific parts ■ Documented analysis that ■ User guides and documentation for determines probability distribution(s) models and frameworks of underlying demand processes ■ Descriptive analysis and validation of stochastic demand processes

Our methodology uses a statistically-determined strategy to gather the most representative sample from the entire data set, while minimizing collection time and effort. Our toolkit contains ready-to-use, detailed data requests. Commercial Airline MRO Example

Data Requests Our toolkit contains data request templates for different parts environments: Wholesale distribution > Retail aftermarket

> Airline MRO Heavy Equipment Wholesale Example > Telecomm maintenance > High-tech spares > Dealer channel > Public utility or energy > Military logistics > Fixed asset maintenance Plant operations

Our sampling techniques allow us to obtain data that statistically represents the broad parts population; typically, we will extract transactional data from each network echelon and location to model the entire supply chain. Sampling Methodology Overview

The sampling methodology not only creates a statistically-valid sample for analysis, it also uses a randomized process to avoid injecting bias into the model or analysis.

Our approach is to select segmentation criteria based on business requirements and objectives; the goal is to create meaningful and unique parts classes that will provide new operational insights. ejagβΛfiϊlSlHSie ^■ ≠ Ss fti

Criticatit¥ of "PP ^er, ^®a t Total replenishment cycle time and variation, from j High, medium, or low, depending upon the rⁱme varⁱatⁱon j _{order p}|_{acement t0} receipt and confirmation. Customer and Duration Demand 5 application of the end user. Degree of Supplier Delivery , Complete order fill rate (COFR), so that all units, To what degree will future sales be lost if the order Performance lines and orders are delivered on promised date. Demand is not filled. Perishability Fart Profitability I Standard Cost of ' Standard acquisition cost or price of the part to the j A measure of the profitability of the part sale to the or Contribution Part i business. business - preferably a net profit measure. Margin Replenishment Average number of replenishment orders made per Economic i Annual sales and profit contribution of the Importance of _F Frequency '. year for the parts. '■ customer, with estimates of future contributions. Customer r ' « Strategic f Economic Lot The size of the optimized replenishment lot size to \ Strategic attractiveness of the customer based on Importance of Size the supplier. ! expected growth, ease of partnering, etc. Customer , Supplier Relationship types, which can range from open Demand Monthly variation in order volumes, typically Relationship Type market buying, to contractual, to strategic. Variability defined as the coefficient of variation (C_v). Interrelation With Part Lifecycle , Lifecycle phases include prototype, introductory, j Impact of gained or loss parts sales on other Other Segment Phase ramp-up, maturity, decline, phase-out, and scrap. I related businesses, such as Service. Businesses

Our methodology creates classifications using a strategically-driven process, with the goal of identifying "logistically distinct businesses," or LDBs, within the broader parts supply chain.

LDB Name: "Churn and Burn" Color Label Italics LDB Characteristics — ■» »• LDB Parts Flow # LDB Parts Deployment

High-level customer segmentation models are also built, primarily to act as input into the cost- service equation used to set service levels for specific parts and customers.

Optimize service levels for the 'Laggards" • Selectively divest volume and rigorously manage the relationship • Offer a standard, no-frills, non- customized menu of services • Implement price penalty for deviations outside of service offerings

Optimizers" • Selectively invest in capabilities to drive volume • Look for low-cost methods of generating new revenue relationships • Reevaluate economic relationship on a regular basis

In addition to customer segmentation models, part segmentation models are also built, which are used to develop customized planning rules. ^ - -- — VMI Decision Framework " Criticality Definition - how a stock-out High VMI program would only be High Poor candidate for VMI; will impact a customer's operation successful with effective use internal DRP system to ," High - causes unit down demand collaboration manage; establish safety Medium - equipment has limited program and time-tested stock parameters and window of operation relationship in place service levels Low - little or no impact Criticality Criticality Excellent candidate for internal replenishment VMI; Business should Unit Usage Definition - a relative approaches such as DRP measure of sales or issuances per time α> establish service are best approach with requirements for vendor to period O Low levels follow Low moderate service Create an index where the fastest CO mover = 100% and index all other High VMI candidate in the long High Poor candidate for VMI; parts to that run after effective supplier optimize single-item / partnership and single-location stocking Usage Variation Definition - a relative confidence has been controls and review measure of how volatile a part's developed performance regularly demand is compared to its volume Criticality Criticality Coefficient of variation (C_v): ratio of Good candidate for ERP the standard deviation and mean Excellent candidate for auto-replenishment; fine- > C_v = (standard deviation / mean) * VMI; establish consensus tune ROP/ROG/S5 100 per time period f replenishment process and parameters over time to > C_v < .25 = smooth demand Low parameters with vendor Low minimize inventories C_v < .50 = erratic demand > C_v > .50 = volatile demand Low Usage Variation High

The segmentation models are critical tools in developing multi-echelon deployment strategies; they are also very effective in graphically conveying these new strategies. _Ψ= _mB^ _{s ss}

Multi-echelon Deployment Strategy

High

Developing the customer segments is the starting point for defining customer service levels - an implicit or explicit approach can be used to define the level for each segment or LDB. Alternative Definitions of Service

Best Fit IMPLICIT METHOD - Using a calculation process to imply the optimum ■ If values for input parameters are available for use service level as a function of stock-out costs and carrying costs. ^a If business has no existing customer service strategy or doesn't understand cost-to-setve concepts EXPLICIT METHOD - Use management expertise to set the acceptable * Explicit methods are valid if management is capable minimum number of planned stock-outs. of assigning discrete, differentiated service levels to customers r Cycle Service Level- Specified probability [P of no stock-outs per replenishment cycle ^a For parts with numerous periods of zero demand ■ For a business where back-orders are common - [P<] - Generally known as an availability measure can be viewed as the probability of no back-orders - Includes the probability of periods having zero demand during the year

Fill Rate Level - Specific fraction of demand [P₂] to be satisfied routinely from ^a Valid for higher-volume parts the shelf - Good fit for distribution system where replenishment - Most common of all measures - assumes no backorders or lost sales lead times are fixed or very costly to change

Ready Rate - Specific fraction of time [P₃] during which net stock is positive ^a Complex to use to determine optimal inventory policy - Least common measure - has application in emergency environments < Applicable for end-of-life products

Our approach recommends using an implicit, quantitative method for critical and high-value items, while using an explicit method for less-critical or less profitable parts.

When a sen ice level is implicitly determined, the planner is considering costs and service to determine the optimum balance for a customer or LDB.

Cost-Service Model

Characteristics of Cost-Service Model _*■ Graph is primarily illustrative - increasing service (e.g., number of parts) is described by a geometric cost function >lf demand is highly fragmented (e.g., widely dispersed asset base) then the cost function becomes exponential at increasing service levels -Rationale for higher service levels is driven by higher stock-out costs (internal or external) ' For feasibility, the model can be applied to groups of similar parts instead of each end item Executing this model requires stochastic

simulation (Monte Carlo) Lower Fill Rate Percentage - probability the desired part Higher will be on the shelf coinciding with its demand event E[D] = Expected annual demand B₁ = Cost of a single stock-out ($) Total Costs = Expected Cost of Stock-outs + Inventory Carrying Costs P_(s) = Probability of a stock-out given s parts (1 - fill rate percentage) Total Costs =

+ hvs h = Holding cost (%) v = Part standard cost ($) s = Number of service parts

Two critical input variables to the cost-service equation are the cost of a stock-out [BJ and annual holding costs [h]. The value of both of these variables is a function of inventory type, location and demand type.

Stock-out Cost Example [BJ Carrying Cost Example [h] Field Service or Carrying Cost Example Maintenance Environment Component Value Description B_{1 =} (LC₀d + b)?p L = Number of units supported by this part Cost of Capital to the 8% Charge the business applies C₀ = Cost per unit per hour of a service outage Business to capital projects d = Expected duration of an outage (hours) b = Cost of expediting to correct an outage Obsolescence 12% Estimate of the annual dollar ?p = Probability that an outage occurs if a part value of parts going obsolete; is not immediately available. For highly a function of their useful life critical parts, ?p = 1.0. Retail Aftermarket Storage and Handling 6% A fully-loaded cost per Business Property Tax square foot calculation for B_{1 =} (LP₀d + b)?p ■ L = Number of customer orders for part during depots stock-out duration = Incremental profit per part sale Property Tax 1.25% State-specific d = Expected duration of stock-out (hours) b = Cost of expediting to correct a stock-out Insurance 2% The business can provide ?p = Probability that purchase is made at this number alternative outlet if part is not available. For non-proprietary parts, ?p = 1.0. 29.25% h = Annual total carrying cost rate as percent of gross inventory investment

Annual Holding Cost = Holding Cost Rate (h) x Cost per Unit x Number of Units

Once carrying costs [h] and the cost of a stockout have been determined [BJ, an explicit service level can now be calculated.

Calculate the safety factor [k] now that stock-out and holding costs have been determined: Demand per year (units) DB₁ Cost of a stock-out (determined previously) Replenishment order quantity (units)

I_* *_** Carrying cost rate (determined previously)

Qhvs_L Standard cost of the part (dollars)

Standard deviation of demand during the replenishment lead time (units)

From the [k] value, now determine the required service level using MS Excel: p[kl = Service level to be supported for a part or customer k = factor determined previously μ_D = Average demand during replenishment lead time p[k] = 1 - NORMDIST μ S_t., TRUE) s_L - Standard deviation of demand during replenishment lead time TRUE - Returns the cumulative distribution function (CDF), not the discrete function

Scientifically quantifying the k factor is the most important factor in enabling the business to meet service levels while maintaining minimum inventory levels.

<* W IO D CO o ..= φ W W OI W W M N W at ⁽M <M iM OI <M W CM W W « n f^{" "}:: CO ; co 13 «- 3 , D) c I 1 ■^Q E » a? o T- « m ^■* U3 to κ o c6 o »- « m _"* « © N © ffl o -o '_ in?^■ 5€* ώ ∞ φ m ώ ώ c^ Kή Φ Φ Φ Φ σ) δ Φ Φ Φ φ φ θ> a) σ> σs a3 <Ji a} θ) a) Φ u o> θ3 c

The safety factor [k] can also be directly calculated using formulas that approximate a cost- service function for normally-distributed demand. These approximations are referred to as "loss functions" that describe the fraction of total quantity ordered but not filled. Determining [k] for a P₁ service measure (cycle service level):

Determining [k] for a P₂ service measure (fill rate level):

The segmentation classes, along with their service levels, are then ready to be assigned to their "best-fit" planning model - a decision driven by the requirements and rules that define that segmentation class. _, . .. . , _{^ Λ}. ^ Planning Model Continuum

Appl

Deployment Models ^τ'^me supply methods Single location, deterministic Multi-location, stochastic Multi-echelon, multi-period (weeks forward demand assumption demand assumption assumptions with stochastic coverage, DOH) demand assumption Replenishment Models Heuristic (informal) lot Static, single-factor optimization Time-phased replenishment Time-phased dynamic lot sizing rules lot sizing (EOQ) in single planning (DRP) with lot size sizing in a multi-echelon location optimization environment Our methodology considers the entire, broad spectrum of planning models - ranging from the most basic to the most advanced approaches. ¹Cost of application would include technology, personnel, training, and overhead costs

Within classical inventory theory, there are two distinct planning models to choose from: the continuous review (or Q) system, and the periodic review (or P) system. Continuous Review (Q) Periodic Review (P)

The inventory position (IP) is reviewed after each transaction The inventory position is reviewed at fixed time intervals, and a reorder decision is evaluated where a new order is placed Sometimes called the reorder point system (ROP) The number of periods (P) between orders is fixed Replenishment frequency can be customized by setting R Lot size (Q) can be customized by setting T, the target level Lot size (Q) is fixed and is usually optimized to meet some Typically results in greater safety stock, but perpetual inventory business constraint: space, price discount, carrying cost systems are not needed

The complicating factor in selecting the best-fit planning model is when the demand for a part is non-normally distributed; in other words, its random, exponential, or exhibits other characteristics that make it difficult to plan. Normal Dlstribittton Mean The Normal distribution is described by two parameters: the mean (central tendency), and the standard deviation (dispersion) Referred to as an "independent distribution," because the mean and the standard deviation are independent of each other Normally-distributed demand can take any value, including negative values Provides a good representation of demand when volumes are moderate to high The Poisson distribution is described by a single parameter [?], which is both the mean and Polsson Distribution variance (the variance is the square of the standard deviation) Mean ^■ Referred to as a "dependent distribution," because the mean and the variance are the same value ^s The Poisson distribution represents a truly randomized demand process, where any demand event in any time period is equally likely to occur J R ⁸ More pronounced if demand sources are widely dispersed and inventories are fragmented ^a Demand is restricted to positive integer values ⁸ Provides a good representation of demand for very low volume and sporadic-demand parts The next step in our methodology accounts for identifying the probability distribution of the demand process of a part - normal and Poisson are only two of several possible candidates.

Another category of planning models includes multi-echelon optimization techniques. Applying these models is only possible through advanced planning systems capable of this type of decision support. Advanced Optimization Techniques Representative Vendor ¹

Accenture has experience implementing and using these packages, and can determine their fitness for a given parts environment. ¹See the Appendix for a discussion on these optimization vendors

A third category of planning models is time-phased replenishment planning (TPRP). This approach can combine transactional replenishment with deployment and lot sizing techniques across a high-volume, multi-echelon parts network. TPRP Structure

Atlanta Central Stocking PDC San Diego Customer Item ID: C01140039 Depot Item Descrip: SS Coupling Ratio Tap 90/10 1x2 Delivery Time: 40 days Internal move Jan Feb Mar Apr May Jun and external 2003 2003 2003 2003 2003 2003 purchase

orders Orlando Customer Depot One TPRP table for each part-location

combination

TPRP is significantly different than the single-item, single-location models discussed earlier; in some applications, there are significant benefits of TPRP over these more basic approaches. TPRP Advantages

TPRP will integrate service demand across multiple field locations, and can be used to synchronize production operations for internal parts The effectiveness of single-item, single-location models is limited due to their "stable demand" assumption - More basic models require updating their parameters (s,S,Q,R) if demand is fundamentally shifting - TPRP can dynamically account for shifting demand and adjust safety stocks TPRP can be used to show planned shipments, allowing for outbound shipment planning and freight optimization TPRP can incorporate demand from all sources: parts forecasts, internal orders, customer orders, and field demand for failed parts

In addition to assigning the best fit planning model, creating the LDBs and part segments also allows the business to align the best fit forecasting model to a part or customer segment. Classes of Forecasting Models

Time Series Models Forecast Model Selection Matrix ^■ Often called "naive" models because of their assumption that future demand is a function of historic demand ^■ Good fit for normally-distributed, higher-volume demand ^■ Examples are weighted average, exponential smoothing, ARIMA, Box-Jenkins Correlative Models ^■ Regression models used to "explain" (forecast) the variability of the dependent variable (demand) as a function of the independent variables (demand drivers) ^■ Good fit if data is obtainable on demand drivers (installed base, failures, weather, promotions) and for slow-movers ^■ Examples include multiple linear regression, econometric models, Cobb-Douglas production function Human Judgment Models ^a Used to override or modify the outputs of statistical methods ^B Good fit if there is no understanding of the demand process or little data to use for reference ^a Examples include the Delphi Technique ^a Common application is hybrid model between judgment and

time series ¹Demand process understanding: knowledge of installed base, data on failures, data on as-is configurations, knowledge of seasonality factors, etc ^aDemand predictability: a combination measure that looks at historical forecast accuracy and demand volatility (C_v)

In this step, we employ a range of statistical tests to identify, or "fit," the demand process to the most likely probability distribution that represents the demand. Distribution-Fitting Process 9 Collect Demand Analyze ata f r Perform Distribution Bun Comparative

^■ Ideally, data is collected for a part segment or LDB to represent multiple, similar SKUs ^■ More data is better - monthly demand for 1-3 years is required ^E A manageable set of candidate demand processes can be selected and these used as proxies across all other, similar demand processes to minimize data needs

Our experience has identified a broad set of demand processes specific to service parts; to truly optimize deployment and replenishment, the demand process must then be matched to its probability distribution.

Incorrectly assuming a specific demand probability distribution when calculating safety stock will result in over- or under-service, depending upon the situation.

T3

SfBiϋi

This phase of our methodology answers two sets of specific questions for each SKU at each location in the network.

- At wholesale or centralized parts depots? - At field or retail outlets? - At customer sites? - In service vehicles? To ensure that: - A specific part is always available when needed? - A part is available 95% of the time? Or 97%? Or 99%?

When should inventory be ordered or moved, and by how much?

> When should inventory be ordered? - As soon as its issued or sold? - When Q units have been used? When the level gets to S? - Every R time period? How much should be ordered? - The same quantity every order? - Enough to last R time periods? - A quantity to satisfy a business rule?

When calculating the TSL, we will use specific policies, such as the single-item, single- location models (S-1 , S) and (s,Q). Both of these models perform well over a wide range of demand volumes and types. Example Control Models Advantages > Both of these models can support a wide range of parts environments - (S-1 ,S) performs very well for low- and sporadic-demand parts - (s,Q) performs very well for moderate- to high-volume parts Demand variability, supply variability and service requirements are all explicitly accounted for >^• The policies are easy to understand and implement, and can be automated in a TPRP system Appropriate Applications These models can be used in either customer-driven demand businesses, or in true maintenance environments - (S-1 ,S) is appropriate for deploying slow-moving items to forward stocking locations (e.g., MRO line station, satellite hub for telecomm network, parts closet for utility service) - (s,Q) is appropriate for higher-volume parts where its desirable that replenishment lot sizes can vary (e.g., wholesale PDC, dealer parts inventory, plant maintenance stockroom)

1. (S-1 ,S) Overview

Input Variables Required 1) ^« Demand or failure rate [?] • Number of parts in service [N] • Supply replenishment time [T]

-4 ■ Variation in replenishment time [s²] • Probability a new part is defective [p]

Policy Features

Policy Description ■ Optimal policy for slow moving, high value inventory ^■ Will determine the TSL needed to meet a desired ^■ Order whenever a unit is used or sold from stock service level ^■ Often called "issue one, replace one," or lot-for-lot ■ Easy to understand and implement (L4L) ^B Can be combined with TPRP for automation ■ Order an amount that returns the inventory level to the order-up-to-level of [S] Policy Applications ■ Order amount is fixed and is typically has a value ■ Where the optimal order quantity is 1 or close to 1 of 1 or near one ^a Where high carrying costs outweigh ordering and transportation costs

1. (S-1,S) Data Requirements

Input Variable Name Description Demand or Field ^■ For maintenance environment: percentage of installed units that fail per time Failure Rate period due to the specific part ^m For distribution business: percentage of total demand that the business will experience per time period for the specific part N Number in For maintenance environment: number of installed units supported by the Service specific spare part For distribution business: number of customers with equipment supported by the specific service part Supply ^■ Average time for the location to be replenished from a supply source - this is Replenishment location-dependent <• Supply source could be internal or external Time

Variation in Supply " Should be calculated using empirical data for the supply source of the specific ^ST - R.eplenishment ^■ ζ C?o^rtmputed , as t .he variance, wh , .ic ,h . is th _.e sum of the squared deviations part used for a repair event will be defective out of the package P Probability a New ^■ Probability that a Part is Defective ^■ Should be calculated using empirical data

1. (S-1 ,S) Algorithms

Review the overall (S-1 ,S) deployment algorithm

S a Order-up-to-tevel p_u = SΛean demand during replenishment lead time S = 1 + μ_L + (s_L) S β Deman variation during reptenϊshment Isad time » Satety factor

Calculate the [k] value

As demonstrated earlier, calculate the k value based on p.,] and [hj. Use th© appropriate solution procedure - for either normal or non-normal demand.

Calculate the [S] value

S =

Note ML and . are algebraically embedded in this algorithm, there Is no need to discretely calculate these input variables

1. (S-1, S) Algorithms

Modify the [S] value, if needed

2. (s,Q) Overview

rical, handling, and [D]

Policy Description ^■ Optimal policy for moderate and higher- volume inventory ^■ Order whenever inventory level reaches the * Will determine the TSL and reorder quantity needed reorder point [s] to meet a desired service level ^■ Often called the Kanban, or 2-bin system ^■ Easy to understand and implement ^■ Can be combined with TPRP for automation ^■ Order a fixed quantity [Q] (e.g., EOQ) ^■ Reorder point [s] is calculated to account for ^! Policy Applications demand and demand variation during ^■ Where the optimal order quantity can be > 1 replenishment lead time ^H Where replenishment lot sizes need to be optimized for business constraints (discounts, space, etc.)

2. (s,Q) Data Requirements

Input Variable Name Description Same as (S-1 ,S) ■ See previous slides s*,P

-4 Unit Replacement * '* purchased, the acquisition cost of the part used by the business in their COGS oe _OS calculation ■ If manufactured internally, the fully-loaded standard unit cost of the part

Inventory Carrying * See previous slides Cost Percentage ^{B Used as a} P^θrcenta9e input

Replenishment * All of the direct and indirect labor, materials, and allocated overhead consumed Order Costs in fulfilling a replenishment order ■ Freight, warehouse labor, clerical labor, etc.

Annual Unit ^a Estimate annual unit demand (sold or consumed) Demand

2. (s,Q) Algorithms

For environments where supplier lead times are very lengthy and/or volatile, the safety stock equation can be modified to account for that duration and variation. This is most commonly applied to external replenishment orders, not internal orders.

Demand Uncertainty Supply Uncertainty

Demand uncertainty is typically expressed using Supply uncertainty is typically expressed using a some measure of dispersion, such as the standard measure of lead time variability, such as the standard deviation of demand, forecast error, or the standard deviation of delivery lead time deviation of forecast errors

For environments where supplier lead times are very lengthy and/or volatile, the safety stock equation can be modified to account for that duration and variation. This is most commonly applied to external replenishment orders, not internal orders, (continued) ROP and Safety Stock Formula

Understand the structure of the revised stocking formula Reorder point =s Demand During Lead Time + Safety Stock

^■ μ_L = Mean demand during mean replenishment Jea ime * mean demand (units) κ mean lead time (time) ^■ s_s = Demand variation over mean lead time = mean lead time (time) x standard deviation of demand (units) ^» s_L = Demand variation during replenishment lead time » mean demand (units) x standard 'deviation of lead f me (time) » = Safety factor Note: For any input values that are negative, square all the terms and take the square root of the final value

In addition to the various deployment models that can be used, our methodology will also select the best-fit replenishment order policy for the business.

Example Lot Sizing Algorithms

As a result of the inventory optimization process, new tools are typically created that can be used by planners, buyers and forecasters to improve their own effectiveness. Representative Tools

New tools for daily execution and process management icfua/ c//^'e/7r model Working software models to make the new used by buyers and planning approaches sustainable planners to plan New analytics and metrics to monitor financial internal and external and operational performance replenishment orders, and to manage slow- Fact-based methods for setting performance moving SKUs targets

Scientific Service Parts Management^® contains Accenture's proprietary tools and methodology for parts inventory optimization; it has delivered hard-dollar benefits for multiple clients. Asset Definition Asset Inventory Accenture's proprietary methodology and tools for Accenture's proprietary methodology and tools optimizing the planning infrastructure in a service Data requests, interview guides parts business Best practices, industry benchmarks Provides specialized analytical approaches that customize planning processes and parameters to Statistical analysis tools for sampling, parts support service levels and operational goals segmentation, inventory analysis, and probabilistic modeling Takes a lifecycle approach to designing, testing and implementing new capabilities to deliver Comprehensive survey of parts planning breakthrough levels of cost and service performance models in the parts business Deployment algorithms to meet a variety of parts demand processes Client Applications (Internal Use Only) Test and deployment procedures Case New Holland (CNH) GE Parts Corning Photonics Technology Division (PTD) Pratt & Whitney Cingular US Airways Pacific Bell

Our strategic diagnostic is an automated toolkit that assesses the degree of alignment between the strategy and execution of a service business.

Asset Definition Asset Inventory A software toolkit used to assess the strategic Creates a comprehensive business process, alignment of a service business technology and organization assessment for a Detailed strategic best practices on service strategy, service business. oe

-4 field logistics, parts management, and customer Driven by an in-depth "strategic best practices" interaction framework across the enterprise. Cross-industry financial benchmark database and Provides detailed financial and shareholder proprietary economic profit calculator value analytics. Business case engine integrated with automated Includes both a financial and operational assessment scorecards benchmark database for OEMs and service businesses. Client Applications (Internal Use Only) Culligan Water Ricoh Family Group Case New Holland (CNH)

To aid our practitioners and clients, Accenture has developed and used several implementation acceleration tools to facilitate better, faster and lower-cost implementations. These assets are part of our Service Parts Management offering. Asset Definition Asset Inventory Delivers a comprehensive application architecture Business case tool set and scorecard and integration requirements for a best-in-class Industry-specific service parts management parts planning capability process model oe oe Contains our program management and project Deliverable templates and samples delivery methodology for design and deployment of the solution Preconfigured methods delivery manager (MDM), fast-start implementation jump-start kit, Contains detailed case study on similar and configuration guidelines and demo implementations Interface design and data mapping for SAP Reduces the time, design, and implementation risk integration of complex projects Training material for templates for service parts Client Applications (Internal Use Only) management processes (i2 Service Parts Planning, Service and Budget Optimizer, and Daimler Chrysler Demand Fulfillment) Applied Materials Siemens Power Generation - Westinghouse Direct TV Harley Davidson Motorcoach

Using our deep, functional knowledge in high-volume parts logistics, Accenture has developed our PEA offering - a toolkit of assets used to design/build these complex capabilities. Asset Definition Asset Inventory Accenture's complete lifecycle program Strategic requirements management methodology for the design, build and ¹ Market intelligence after sales / parts deploy phases of a high-volume parts logistics ^B State-of-the art distribution systems & channels solution Functional solutions Contains industry-specific content on -' Process models and best practices process/technical architecture, integration ⁿ System capabilities, design/build methodology requirements, and deployment programs Technical architecture Client-tested in multiple parts environments Development, execution, operations architecture Training and communication - Training and communication content Client Applications (Internal Use Only) Rollout Kit '^• Implementations concepts and tools for each phase

Accenture's MRO offering has helped our maintenance clients optimize their core processes, technology infrastructure, supply chain, and organization design.

Asset Definition Asset Inventory Accenture's comprehensive suite of offerings all MRO 3-day Financial Assessment - A developed to optimize and reengineer the MRO structured, in-depth analysis of an MRO organization business used to identify value targets Has been developed and implemented at multiple MRO Process Model - Uses cross-industry clients across different maintenance environments experience and commercial best practices to Contains specialized, client-tested assets across the present a comprehensive process model for dimensions of technology, process and organization design and assessment purposes. Provides a point-of-view on MRO-specific MRO Supply Chain Assessment - A value requirements for CRM, supply chain, back-office, targeting exercise focused on MRO supply strategic management, product management, etc. chain areas, supported by a range of assets such as business cases, analytics, metrics, Client Applications (Internal Use Only) interview guides, and assessment tools

Accenture has developed extensive parts-specific thought leadership through our client work and participation in industry events and committees.

Asset Definition Asset Inventory (sample) Accenture's global supply chain practice has "Making Collaborative Planning Work in the developed thought leadership and specialized Real World: A Case Study of Harley-Davidson assets that are focused on service parts and Its Dealers." APICS Conference, Doug management Derrick, 2002. The thought leadership has been developed as "Service Management: Building Profits After the published articles, white papers, and conference Sale." Accenture Supply Chain Perspectives presentations White Paper. Mike Dennis, Dr. Ajit Kambil, Currently, we have Accenture executives that are 2003. members of the APICS Special Industry Group (SIG) "The Forgotten Supply Chain: How OEMs Can on Service Parts and Remanufacturing Escape the Value Squeeze by Optimizing the Our Knowledge Xchange^® is a global repository of Aftermarket Business." APICS The thought leadership and client work that contains our Performance Advantage, Doug Derrick and best thinking on service parts management Robert Jacoby, 2003. "Extending the Service Efficiency Frontier with New Service Management Models." Accenture Outlook and ASCET Volume 2, Mike Dennis and Adam Wolf, 2002. "Increasing Customer Loyalty Through Service Parts Management." ASCET Volume 1 , Tom Jenkins, 1999.

For the Service Management team, our clients are our best credentials - we have worked with the global leaders across industries to deliver service and parts excellence.

Optimizing the aftermarket parts supply chain by reengineering supply, demand, and inventory core processes

Business Challenges Results ^a As dealer demand grew, inventories also were growing - but turns => A comprehensive report was produced that were dropping and were significantly less than industry averages documented $47-60M in operational cost ^a Dealers were holding less inventory and placing more emergency savings (15-20%), and $40-60M in orders; the dealer network was fragmented, with many small inventory reductions (10-12%) dealers placing frequent orders ³ The findings were produced in collaboration ^a Obsolete and overstock parts were estimated to be 20-25% of with the client and validated by their gross inventories Finance team a The global aftermarket business appeared to be at a cost = The assessment produced detailed models, disadvantage (sourcing and direct labor) and lacked key analytics and benchmarks in the following operational capabilities areas: New safety stock models showing Approach and Solution reductions from improved forecasting ³ Conduct a comprehensive capability assessment of the global Direct labor FTE modeling at depots supply chain of the aftermarket business " Deployment and replenishment ^α Examine core processes and technologies on sourcing, optimization models procurement, inventory planning, demand management, > All-time buy optimization models transportation, depot operations and network strategy • Final report was a time-phase, fully-detailed ⁿ Conduct the assessment across two continents, interview 60+ 5-year master plan with capital budget personnel, conduct site visits, and perform benchmarking ^j Host executive workshops to collaborate on assessment findings Accenture Contact and recommendations ^a Robert Jacoby

Business Challenges Results ■ Commercial airline carrier was trailing industry cost and service 3 In a 5-week capability assessment, the benchmarks in its MRO business (maintenance, repair and team identified $170-200M in parts overhaul) reductions (31-38%) across rotables, ^a The organization did not have a history of supply chain or repairables and expendables, and a materials management expertise positive cash flow impact of $61-76M * Planners, forecasters and buyers were relying on legacy systems ^α For each MRO supply chain function, and outdated procedures to manage the parts supply chain detailed analytics and recommendations a Other core processes (tooling management, distribution) were were developed ineffective and not supporting enterprise service objectives = A time-phased and prioritized list of 8 projects was created to realize benefits ^ Accenture entered into a gain-sharing Approach and Solution arrangement for the next phase 3 Conduct a fast-paced assessment on all core processes of the ³ The inventory optimization phase resulted MRO supply chain: procurement, demand management, supply in a 30% reduction in pilot inventories with management, tooling management, surplusiπg, an incremental $20M in process distribution/fulfillment, network strategy, materials control ¹ System-wide availability was maintained at = Conduct site visits and interviews, perform benchmarking and 95% extract data from business systems for extensive analysis < 3-10% of procurement spend on parts was - Review existing IT applications and technical infrastructure for avoided effectiveness Accenture Contact ^u Develop working model of potential inventory reductions T Robert Jacoby

Business Challenges Results = OEM was faced with a dealer-dominant channel, where a portion ^τ Participating dealer can carry a larger of the dealers lacked fundamental supply chain and retail inventory assortment while improving retail availability management skills of parts & accessories by 14% and general ^B The number of product lines and products offered had grown merchandise by 19% steadily over the past 10 years Inventory turns for P&A increased by 11 % ■ Dealers worked in a capital-scarce environment, making profitable and GM increased by 4% inventory investments very critical ¹ Wholesale sales growth rate for pilot ^a Dealer and OEM technologies were very basic and did not lend dealers was 6.4% greater than the themselves to channel inventory optimization wholesale sales growth rate of he dealer network for P&A, and 15% greater than the dealer network for GM Approach and Solution ³ Develop a collaborative inventory management solution, consisting of a process, organization and technical infrastructure ^ Develop the process such that the solution can be configured for the unique market needs and requirements of each dealer -¹ Begin with a low-tech pilot concept to prove the operating model and deliver initial benefits ¹ Deploy a technology pilot at selected dealerships to fine-tune and Accenture Contact stabilize the product ¹ Doug Derrick ^α Deploy across the broader dealer network

Business Challenges Results * Business was booming - revenues were expected to triple in a > .The team identified $22-25M in inventory two-year period for this fledgling business improvements from process, organization ■ Inventory accuracy was below 70% and unplanne stock-outs and technology changes were harming customer service in a competitive industry , ^a A fully-detailed capital investment plan was » Internal processes were broken - basic inventory controls, supplier built as a supply chain roadmap management, forecasting, and performance measurement , ^B 10 separate improvement projects were ■ Order fill rates and ship-to dates f om external an internal identified and planned suppliers were being met on average less than 35% of the time _. ^s The primary project to build a decision ^m Planners had no tools or visibility to analyze, material performance support tool for buyers/planners was launched: Tool conducted automated materials Approach and Solution segmentation ^a Conduct an exhaustive 8-week root-cause analysis of the parts Aligned segments to optimized supply chain for the business planning models α Develop and refine a comprehensive list of improvement > Calculated optimized deployment and hypotheses replenishment targets ^B Examine all processes driving parts inventory performance: supply Automated scorecard of inventory management, integrated planning, production operations, metrics and early-warning indicators inventory control, demand management and order fulfillment ^α Conduct detailed benchmarking, analysis and modeling using Accenture Contact large amounts of transactional data from business systems ^α Robert Jacoby

Business Challenges Results * Client had grown in a highly regulated environment that rewarded ^E By realizing a 2.5% overall reduction in service reliability and not competitive agility spare inventory levels, this joint effort has * Inventory was estimated to be in excess of 5% of total inventories saved the client over $1 OOM in capital costs on a $2.6B base via reduced plug-in purchases ■ There were more than a million inventory moves each year among ^α Client has reduced inventories from 7.5 to 715 central office switching centers 2.5 spares per 100 in use, exceeding ■ Deregulation forced client to maximize operating efficiency while industry benchmarks maintaining a high level of service Bar coding and mobile device infrastructure enables client to keep flawless account of the 27,000 different plugs it uses ^■> Project won the 1997 RIT/USA Today Approach and Solution Quality Cup for the service industry ° Accenture and the client developed a gain sharing arrangement - The integrated solution allows users to where our fees were tied to hard dollar cost savings at the client locate plug-ins throughout the entire - Using the principles of scientific service parts management , the company, optimize sparing levels for every team developed a comprehensive parts planning strategy across location, replace defective plugs and report the distribution network on defective history at each location, and ^π Segmentation led to the deployment of specific deployment automatically send replenishment orders to policies and control models for different parts and locations the warehouse or the vendor system ^B Detailed statistical analysis generated new values for deployment Accenture Contact and replenishment execution Robert Jacoby

Results ^s Reduced inventory over $120M in Aftermarket and Spares ^π Increased inventory turns in Aftermarket to 2.14, well above the goal of 2.03 " Conducted logistics network assessment highlighting annual savings of $10-$15M Completed APO prototype ^J Developed order management and inventory management processes

Develop consistent reports, using SAP-BW Approach and Solution and developed standard metrics ^D Accenture entered into a gain sharing relationship with the Client, ¹ Developed design logic for parts ordering where our payment is linked to the inventory benefits realized hierarchy tool ^s Accenture is helping design and deploy a number of tools to " Performed data and inventory analysis to support the inventory optimization effort: identify opportunity areas for further inventory reduction f Training and Change Management - deploy the virtual inventory toolset and new Materials Management reports Received gain-share fees of well over $1M by exceeding joint goals > Materials Management Best Practices - deploy inventory best practices across the Spares and Aftermarket business Accenture Contact > Metrics to track reduction - implement inventory metrics dashboard Michael J. Dennis

Business Challenges Results Inventory levels had accumulated to a level significantly higher ³ Up to a 50% reduction in parts inventory than what was required to support demand and service levels requirements due to optimization of target ■ 60% of materials were spot buy vs. contract procurement levels and replenishment ■ Growing excess and obsolete spare parts inventory ^B Reduced expediting of materials/orders, ^B Procurement was driven by 'buy lists' provided by account teams reduction of excess and obsolete by $7M

© © annually through better planning processes 80% improvement of spare parts forecasting accuracy -^J $20M reduction of purchasing/material costs by converting to blanket PO's based on forecast rather than spot buys Approach and Solution ^j Improved customer service levels: ^m The Spares division is responsible for supplying consumables and capability to deliver 'downs' parts 93% of spare parts for the repair, maintenance an overhaul of their the time within 4 hours, 95% within 24 customer's wafer fabrication systems hours and 98.5% within 48 hours ^c Organization is a reactive culture focusing on expediting to fulfill Improved fill rates due to having the right demand and perceived demand part at the right time, in the right place Program was focused on delivering a comprehensive solution to ⁵ Automated customer order sourcing to improve order management, forecasting, inventory management quickly locate spare parts - Develop regional DCs for fast moving parts and one NDC for slow Accenture Contact moving materials ^~> Duane H. Laun

Business Challenges Results ^s To create an integrated supply chain strategy with processes that ³ Migration of the business to a hub and support the organizations need to reduce cost and improve spoke materials distribution network for the customer service service parts business ^■ Utility had accumulated high material and operating costs due to ^«* Development of a solution that integrates excessive inventory levels and expansive distribution network both demand planning and service parts * Manage inventory across significant number of remote stocking forecasting locations Initiated a Strategic Sourcing program to rationalize the parts supplier base -¹ Reduced spare parts inventories by 40% ¹ Increased crew productivity by 8-12% Approach and Solution ¹ Annual productivity savings of between $3- $5M ^D One of the five electric utilities that belong to the Southern ¹ Strategic Sourcing delivered $4.5-6.8M in Company, one of the largest power producers in the world annual savings * $3.2B electric utility serving central and southern Alabama with 1.3M industrial, commercial and residential customers Accenture Contact ⁱ Performed a comprehensive diagnosis of current supply chain processes for effectiveness and efficiency Erik A. Olson - Developed a network model to test alternative network scenarios and a conceptual design of the new operating model ^ Piloted, tested, and implemented recommendations

Scientific Service Parts Management^® is administered by the Service Management team. Please contact a member of the team below if you have questions or a client application. Key Contacts

-,! * 1 ji. —

©

I -i. ^*

Service parts planning solutions range from specialized vendors to SCM and ERP vendors, which have expanded traditional functionality to support the unique demands of service parts.

ue of Innovation.

http://www.sap.com Germany

Overview Founded in 1972, SAP is the largest enterprise solutions vendor in the market. They have provided business solutions for all types of industries and for every major market for the last 30 years.

©

-4 Headquartered in Walldorf, Germany, SAP is the world's largest inter-enterprise software company, and the world's third-largest independent software supplier overall. SAP employs over 28,900 people in more than 50 countries. The company boasts 12 million users, 60,100 installations, 1,500 partners, and 23 industry solutions. SAP's primary product offering is the mySAP Business Suite, which is the broad suite of business applications they sell across enterprise functions. The SAP APO 4.0 module (advanced planning and optimization) is the planning component to the SAP SCM 4.0 solution. The APO solution has been deployed for service parts environments.

Selected Customers (parts solutions): ■ Porsche ■ Pratt & Whitney ■ VW/ Audi

Key Modules and Solutions: Potential Differentiators: ■ SAP APO 4.0: ■ Ease of integration to the broad suite of SAP enterprise applications ■ Demand Planning ■ Predictable upgrade path ■ Supply Planning ■ Currently engaged in a consortium effort with Caterpillar and Ford to ■ Supply Chain Cockpit build the next-generation parts solution ■ Production Planning & Scheduling

Application Summary With SAP APO, SAP has combined the ERP execution power of the SAP The SAP Advanced Planner and Optimizer builds on the SAP Business R/3 System with advanced data analysis and supply chain management Framework to improve information flow by incorporating real-time

© oe tools. Because SAP has built a robust integration layer between SAP APO collaborative decision support, advanced planning, and optimization into and the underlying execution system, SAP APO can gain immediate and the SAP R/3 System. SAP APO uses a powerful memory resident seamless access to Online Transaction Processing (OLTP) business data. analytical engine and highly specialized, highly configurable data objects While the data objects contained within SAP APO are, in most instances, that offer major new components: structurally optimized instances of OLTP data, they remain synchronized through a series of real-time triggers and messaging, a task that is - Supply Chain Cockpit seamlessly accomplished through the integration services of the Business - Demand Planning Framework. SAP refers to this technique as "semantic synchronization". - Supply Network Planning - Production Planning and Detailed Scheduling The SAP APO server also integrates with the SAP Business Information - Global Available-to-Promise Warehouse using this same mechanism providing access to vital business - Supply Chain Cockpit decision data. Advanced Optimization Techniques and Technology In addition to highly specialized data objects, SAP APO uses a library of The Demand Planning component is a toolkit of statistical forecasting advanced optimization algorithms and a high performance, memory techniques and demand planning features that helps you create accurate resident data processor to perform planning and optimization. You can forecasts and plans. Demand Planning gives you a sound understanding configure SAP APO to provide task-specific, industry-specific, and of all the factors that affect demand, delivering context-based demand company-specific optimization, automated decision, and real-time event planning, which raises forecasting to a new level of sophistication and notification to the under-lying business processes. accuracy. You can analyze actual demand using a variety of tools, such as multiple linear reqression, and incorporate causal factors like orice.

X E L U S Employees: 120 Headquarters: Fairport, NY http://www.xelus.com

Overview The company (formerly LPA Software) provides software that companies use to manage enterprise servⁱce operatⁱons. Its products track contracts and parts, schedule workers, and manage repair center, maintenance, and field service operations, including applications for asset management and recovery. Xelus also provides consulting, training, and other services. Customers include 3Com Cⁱsco Systems, and Dell. Xelus has expanded beyond its roots in the telecommunications industry ⁱnto fⁱelds ⁱncluding aerospace and defense, transportation, and utilities. Founded in 1972 the company's investors include MMC Capital and Oak Investment Partners

Selected Customers (parts solutions): Siemens SBC Cisco Systems Verizon IBM Honeywell Samsung Dell Xerox 3Com BAE Systems British Airways Subaru Compaq Sun Microsystems TRW Aeronautical Systems Hewlett Packard Lockheed Martin

Sales: €125.7 million (2002) Company Type: Private

of [anttøtvisctlon. Employees: unknown Headquarters: Milan, Italy http://www.finmatica.com

Overview Finmatica acquired Mercia Software in June 2002 FINMATICA is an Independent Software Vendor (ISV) which develops products and solutions for Supply Chain Management, Finance, Information & Internet Security, Document Management and Transportation Management.

FINMATICA has recently acquired Mercia, a leading provider of advanced planning systems for service parts businesses.

Selected Customers (parts solutions): ■ Saturn ■ Canon ■ GM ■ Honeywell ■ Daewoo ■ Rolls Royce ■ TRW

oπware

Key Modules and Solutions: Potential Differentiators:

FINeCHAIN suite:

■ Demand Planning ■ Develop leading-edge practices for the forecasting of slow-moving

■ Strategic Inventory Planning and sporadic demand parts

■ Distribution Requirements Planning ■ Positions itself as the leader in forecasting for stochastic service These modules are part of MerciaLincs parts

Application Summary Finmatica acquired Mercia Software in June 2002

MerciaLincs allows organizations to take control of demand in order to plan inventory, distribution and requirements in the most efficient manner. By using a combination of sophisticated, low intervention statistics and subjective input, MerciaLincs demand planning easily handles difficult situations such as promotions, new product launches, obsolescence, etc. By balancing and optimizing inventory throughout the network, maximum customer service can be delivered for the minimum investment in stock.

SERYlGi πCS ^{SaleS: <$10 million} (^est> Company Type: Private Employees: <150 Headquarters: Atlanta, GA http://www.serviqistics.com

Overview Servigistics is a provider of global service parts management solutions specifically designed to decrease costs, increase profitability and improve customer loyalty by streamlining and automating complex service parts operations. Servigistics solutions have been tested and proven by leading global companies, including Axcelis, Cray, Dell, EMC, IKON Office Solutions, Subaru, Toshiba Medical Systems and UPS.

Servigistics mission is to help leading companies rapidly improve customer loyalty and profitability through superior service parts management.

Selected Customers (parts solutions): ■ AGFA ■ Dell Toshiba Medical Systems ■ Akibia ■ EMC UPS ■ AUSPEX ■ Ikon ■ Axcelis ■ Patterson Dental Company ■ CNT ■ River Stone Networks ■ Cray ■ Subaru

Key Modules and Solutions: Potential Differentiators: ■ Servigistics Location Master ■ Completely web-enabled, J2EE architecture - allows for globally- ■ Servigistics Lifecycle Manager distributed planning and collaboration ■ Servigistics Profit Analyzer ■ Impressive install base with a product that is gaining market momentum

Application Summary Servigistics 7.1 is the optimal solution to deliver your service Easy to use, the global solution delivers the key functionality parts strategy, making certain you have the right part, at the required to support your core businesses processes and right place, at the right time. complex service parts requirements, including: Our proven and tested Service Parts Management solution Sophisticated Optimization Logic - plans the right stock, was built from the ground up with one goal in mind - to deliver at every echelon and every location, from central to field beyond expectations and help leading companies like yours rapidly improve customer loyalty and profitability through: What-if Analysis - proactively estimates the impact and cost of various service strategies and decisions before they Information Consolidation - seamlessly integrates are made. information from multiple, world-wide transaction systems into a single decision support system Demand & Lead Time Forecasting - considers seasonality, multiple demand streams and leading indicators Inventory Optimization - advanced optimization to deliver highly accurate forecast science replaces manual processes to deliver optimal inventory plans to transaction systems Complex Parts Chaining & Alternate Parts Planning - incorporated into all processing and decision support to Event Management - configurable logic enables users to maximize inventory options proactively identify important events before they turn into critical problems Exception Analysis - focuses users on critical business issues Global Visibility & Analytics - powerful functionality provides complete parts visibility and analytics to discover, adjust and Executive Dashboard - delivers enhanced management manage key processes and exceptions on a local or global and control via hierarchical views, detailed drill downs and basis performance analyses

http://www.manu.com

Overview Manugistics Service & Parts Management describes its value propositions as: manage your service business as a profit center; provide highest levels of customer service at the lowest costs, while optimizing asset uptime; improve utilization and capacity of operational and production assets; cut service parts inventories, service costs, and purchasing overhead costs; improve forecasting and deployment of service parts and resources; provide inventory visibility across service parts stocking locations; profitably manage part allocations between production and service; and, optimize the complete asset repair loop

Manugistics Service & Parts Management solution provides the integrated set of capabilities needed to effectively manage service and parts operations. Features such as statistical forecasting, inventory optimization, and multi-site finite capacity scheduling help ensure your ability to optimally deploy spares, consumables, and repair resources to meet the needs of both planned and unplanned activities.

Selected Customers (parts solutions): ■ Boeing ■ Defense Logistics Agency ■ Harley Davidson ■ Mitsubishi Motors

Key Modules and Solutions: Potential Differentiators: Service & Parts Management NetWORKS suite of applications ■ Long, relatively successful legacy of supply chain management and (formerly WDS) Traditional SCM modules: planning applications - Demand ■ Recent research suggests Manugistics is one of the most successful - Fulfill - Collaborate vendors in delivering true ROI for their clients - etc. ■ Multiple industry solutions could be leveraged for parts solution

Application Summary Service network planning: The service network is designed to determine Service delivery management: To ensure on-time deliveries, parts plans can be locations of repair parts, field technicians, repair depots, central inventory executed, purchase orders created, and shipments monitored. At the same time, locations, and field stocking locations, in order to balance customer service fulfillment of service calls is based on optimizing service technician dispatch goals and asset uptime with associated costs. queues, technician skill sets, customer priority, contractual response time commitments, and customer location. Service and parts pricing: Pricing for service technician labor, service parts, and service agreements is determined in order to help maximize profitability Reverse logistics: Inbound assets must be forecast and tracked, whether they are for each market segment being serviced. from planned or unplanned repairs, overhauls, customer returns, warranty services, damage, recalls, or other reasons. Repair demand and parts planning: Many factors, including usage patterns, buy versus repair estimates, failure rates, maintenance and repair Collaboration and integration: Service part procurement planning and execution schedules, mean time between failure (MTBF), and mean time between requires supplier collaboration; contract management and service support requires unplanned repair (MTBUR) are used to create an optimal plan for preventive customer collaboration. Manugistics provides an open architecture to support such and overhaul maintenance. internal and external integration. Allocation management: To meet customer priorities and improve overall Event management and analysis: Time-phased event management and critical margins, parts must be allocated between service and production areas, path resolution pre-empt problems. Analytical reporting by part, asset, and while parts and labor must be allocated between the factory service shop commodity and network-wide visibility of parts, materials, and repair and overhaul and field service. schedules proactively alerts maintenance and service personnel about disconnects in capacity, parts, tools, and materials demand and supply — and provides Resource scheduling: Planned work and forecasted non-planned work, resolution paths. scheduled according to simultaneous consideration of resource availability and parts availability, helps ensure alignment of asset planned downtime, Sourcing, content, and knowledge management: Management of service parts capacity, repair and overhaul time needed, labor availability, technician lists, specifications, bulletins, illustrated parts catalogs, and supply sources to capability, parts, materials, tools, customer priorities, and contractual service support service technicians and purchasing allows companies to capture asset 'as- obligations. maintained' histories, planned resource and parts needs for asset repair types, and

Employees: 4,800 Headquarters: Dallas, TX http://www.i2.com

Overview i2 Service and Parts Management is designed to enable enterprises to maximize the utilization of parts, people, budgets and facilities so that they can attain key performance objectives such as high customer service, market leadership, low operating costs and profitability Capabilities include: - Optimize inventory levels using proven parts criticality methodology. - Optimize the tradeoffs between inventory levels, budget and customer service level targets - Determine the optimum mix of parts while considering storage constraints. - Enable service organizations to successfully and profitably service more customers per day. - Respond more effectively to customer demand with the one solution that integrates planning, order execution, and logistics i2 differentiators include: • Solutions deployed in a modular, independent fashion • Functionality tailored for service parts planning • Forecasting techniques • Replenishment planning algorithms • Optimization of stocking levels and re-order points using mixed integer linear programming • Proven installations realizing quantifiable value Selected Customers (parts solutions): ■ DirectTV ■ Southwest Airlines ■ Toyota Motor Sales

Key Modules and Solutions: Potential Differentiators: ■ Service Parts Planning (SPP) ■ Service Budget Optimizer ■ Aggressively positioning themselves as the best-of-breed for SPP (SBO) vendors ■ Have developed a module that specifically addresses the issue of spend and budgeting optimization for a parts business

Application Summary i2's service parts product offerings are supported by two applications: i2 Service Budget Optimizer can help companies: Service Parts Planner (SPP) and Service Budget Optimizer (SBO). - Determine optimized inventory levels for customer service targets i2 Service Parts Planner can help companies: - Rebalance inventories - Recommend replenishment policies at a part location level - Accurately model complex service chains - Improve forecast accuracy Service Budget Optimizer maximizes return on service assets, provides a - Minimize inventory required to meet customer service targets quantitative basis for determining budget and inventory deployment across - Efficiently manage large volumes of parts a service parts network, and can be integrated with tactical customer interaction management planning tools. Many planning tools available today can only support the simple situation of a single stocking location, which is inadequate for large Using Service Budget Optimizer planners can view existing inventory at all service organizations with multiple stocking locations and a complex locations and make inventory decisions based on budget and customer supply network. constraints. Service Parts Planner forecasts demand for spare parts based on historical data and equipment usage and failure rates, sets parts target inventory levels, and generates replenishment plans that consider the forecast for repairable and returned parts while considering alternate parts.

1 1 IV- Sales: Unknown Company Type: Private Employees: Unknown Headquarters: Philadelphia, PA http://www.mcasolutions.com

Overview MCA's Service Planning & Optimization solution has earned a prestigious National Science Foundation (NSF) Innovation Award. The NSF has described MCA's technology as "one of the best examples of research that has influenced practice." Recently the NSF further substantiated MCA's breakthrough developments by awarding MCA its second NSF Innovation Award.

Service Planning & Optimization (SPO™) Product Suite: MCA Solutions' SPO suite is the only solution on the market that enables you to optimize resources in a multi-echelon service supply chain network. Its integrated, web-centric modules were designed with 25 years of deep industry expertise to address all of the challenges you face every day.

Selected Customers (parts solutions): ■ IBM ■ Cisco ■ Boeing

Key Modules and Solutions: Potential Differentiators: SPO Strategy: SPO Tactics: ■ Positioning themselves as the "only" ISV that can deliver true multi- ■ Contract Management ■ Allocation Optimizer echelon inventory optimization across large-scale network ■ Forecast Management ■ Replenishment Optimizer ■ Multi-Echelon Planning ■ Transshipment Optimizer ■ Customer Collaboration and ■ Supplier Collaboration Provisioning Application Summary Service - A Different Kind of Supply Chain SPO Strategy SPO is designed specifically to handle the complex material flows in SPO Strategy enables you to achieve true service supply chain optimization a service supply chain. While the linear models inherent in ERP and because it's focused on systems, not parts. Even across multiple echelons, SCM systems oversimplify the problem, MCA's SPO solution is built SPO models all decisions and trade-offs that impact customer service, to accommodate complex supply chain scenarios, such as inventory investment and logistics costs. Its flexible what-if capability enables remanufacturing and redeploying defective parts or recycling obsolete planners and managers to simulate multiple scenarios, giving them ultimate parts to maintain service levels. Even when traditional ERP software control over all decisions. As a result, you enjoy coordinated optimal stocking suppliers try to adapt their solutions to address service supply chain across your entire service value chain, which in turn lowers the investment complexities, they base calculations on product-side logic, which fails required to meet service goals. to yield networks that are truly optimized for high-quality service. SPO Tactics Service Supply Chain - Optimized SPO Tactics takes the time-phased forecasts and optimal inventory SPO provides true optimization across the entire service value chain. positioning from SPO Strategy and enables your planners to adjust the timing It takes a life-cycle approach, providing for both strategic resource and quantity of repair and new buy orders. Like never before, you'll be able to deployment and intelligent tactical redeployment. SPO can help you source, replenish and ration your resources with extreme confidence, leading to redesign your entire service supply chain network, and to manage to immediate savings as your service parts inventory is reduced. You'll also your service supply chain on an ongoing basis. You will see collaborate more effectively with suppliers to manage for cost and time breakthrough results, as you maintain resources and infrastructure savings. With SPO Strategy and Tactics working in concert, you'll do more and provide uninterrupted service to customers at the lowest possible than design for serviceability. You'll actually execute effectively on a daily cost. basis to deliver better service to your customers.

http://www.bvbaxter.com

Overview Baxter Planning Systems, Inc. was founded in 1993 to fill the gap for planning and logistical solutions specifically targeted to the service parts industry. The key is Baxter Planning System's aftermarket service parts planning and logistics software. Our flagship product, Prophet by Baxter, is a suite of applications that quickly helps companies turn their service parts organization from a cost driver into a profit center.

Baxter's solution has the following exclusive features: • Cost-Based Planning: Our solution calculates the true cost of stocking each part, including "hidden costs" like the cost of stockouts. • Profit-Centered Allocation: Prophet by Baxter gives you the tools to replenish stock based on the actual profitability of each part in each location. • Advanced Product Forecasting: Baxter allows you to map stocking levels to product lifecycles so you don't stock parts that are past their useful life. • Lower Cost of Ownership: Because Baxter offers a hosted solution there are no upgrades to buy, no software to maintain and no drain on IT resources Selected Customers (parts solutions): ■ Agilent ■ Enterasys Philips Medical ■ Lucent Technologies ■ Hewlett-Packard Silicon Graphics ■ Radiant ■ Network Appliance Sony

Key Modules and Solutions: Potential Differentiators: Site Planning Material Forecasting ■ Positioning themselves as a niche provider for aftermarket Distribution Planning Supply Planning businesses, as opposed to true maintenance environment Product Forecasting Support Planning ■ Provides decision support that maximizes aftermarket profitability by Inventory Asset Tracking optimizing costs-service as it drives profitability, not just service

Application Summary Prophet by Baxter™, an integrated software suite, is a planning Distribution Planning solution focusing on the unique complexities of the after-market, ^■ International Site Planning optimizes placement of warehouses and service, repair, and spare parts industries. With increased customer stocking locations. demands for a quick response, often within hours, manufacturers and ^■ Optimizations between stockouts and inventory ensure the highest service organizations must have planning tools that enable a level of customer satisfaction. proactive response to changes in customer requirements and/or ^■ Extensive modeling of the part, product, customer and location material availability. Prophet by Baxter translates forecasts, demand environment ensures inventory is in the optimal location. and material availability into what to buy or repair, when to buy it or ^■ Optimized replenishment/ redeployment ensures response times are repair it, and where to have it in order to meet demand. Prophet by covered. Baxter also optimizes where actual warehouses and stocking locations should be located, based on service contracts and required Supply Planning response times. ^■ End of support life planning predicts material requirements. ^■ Lifetime buy analysis ensures future requirements are met. Forecasting ^■ Supply sourcing decisions such as buy, internal repair, vendor repair, ^■ Product-based forecasting, utilizing product life curves, minimizes and/or whole unit teardown provide for least cost sourcing. backlog and excess. ^■ Extensive trending and adjustment logic increases forecast accuracy. ^■ New product introduction ensures no lapse in coverage.

Our best practices package is one component of our capability assessment, and is a comprehensive tool used to examine the entire parts supply chain. Best Practices Framework

Publications of Interest

1. "An Introduction to Statistical Methods and Data Analysis tøngrtecker, Michael T.; Ott, Lyrnan. Brooks Cole, 5^th Edition. December, 2000. 1 ,184 pgs. 2. "Factory Physics." Hopp, Wallace J; Spearman, Mark L Irwin McGraw-Hill, 2^nd Edition. 2000. 720 gs. 3. "Foundations of Inventory Management" Zipkin, Paul H. Irwin McGraw-Hill. 2000. 514 pgs.

-4 4. "Integrated Inventory Management." Bernard, Paul, Oliver Wight Manufacturing Series, John Wiley & Sons, Inc. 1999, 611 pgs, 5. "Inventory Management and Production Planning and Scheduling Peterson, Rein; Pyke, David F.; Silver, Edward A. John Wiley & Sons, Inc., 3^rd Edition, 1998, 754 pgs. 6. "Manufacturing Planning and Control Systems." Berry, William L; Vollmann, Thomas E.; Whybark, David C. McGraw-Hill Publishing, 4^th Edition. 1997, 896 pgs. 7. Operations Management: Strategy and Analysis." rajewski, Lee J.; Ritzman, Larry P. Addison- Wesley Publishing Company, 2^nd Edition. 1990. 871 pgs. 8. Operations Research in Production Planning, Scheduling, and Inventory Control." Johnson, Lynwood A.; Montgomery, Douglas C. Georgia Institute of Technology. John Wiley & Sons, Inc. 1974. 525 pgs.

Claims

Claims:

1. A method for optimizing TSLs ("target stock levels") for an inventory of stock keeping units ("SKU"), the method comprising: developing a service strategy for assigning a service level for each of the plurality of segments; analyzing the plurality of segments and the service levels for the plurality of segments for identifying at least one logistically distinct business ("LDB"); assigning each of the plurality of segments to a "best-fit" planning model for indicating deployment, replenishment, forecasting and review characteristics for the segments; and identifying a probability distribution function ("PDF") for estimating a demand process of each of the plurality of segments.

2. The method from claim 1 , further comprising calculating a target stock level ("TSL") for the plurality of segments.

3. The method from claim 1 or 2, further comprising calculating an inventory baseline for understanding information about the currently held inventory.

4. The method from claim 3, wherein the step of calculating the inventory baseline comprises generating statistically valid samples to approximate an actual inventory baseline.

5. The method from claim 3 or 4, wherein the step of calculating the inventory baseline comprises running a plurality of data requests pre-defined by industry type.

6. The method from any preceding claim, wherein the step of developing a service strategy comprises segmenting by customer.

7. The method from any preceding claim, wherein the step of developing a service strategy comprises segmenting by SKU.

8. The method from any preceding claim, wherein the step of developing a service strategy comprises segmenting by customer grouping.

9. The method from any preceding claim, wherein the step of developing a service strategy comprises segmenting by strategic attractiveness.

10. The method from any preceding claim, wherein the step of developing a service strategy comprises segmenting by channel type.

11. The method from any preceding claim, wherein the step of quantifying the service level comprises analyzing empirical data.

12. The method from any preceding claim, further comprising performing additional qualitative or quantitative analysis for assigning best fit planning models.

13. The method from any preceding claim, wherein the step of identifying a probability distribution comprises collecting demand data, analyzing the demand data for insights, performing distribution fitting tests, and running comparative analytics.

14. A method for meeting a service level at a probable minimum cost, the method comprising: identifying a probability distribution function ("PDF") of a demand process; running a simulation to test the PDF against historical usage data to determine a service level; adjusting a deployment algorithm based on results of the simulation; and repeating the steps of identifying, running and adjusting until an acceptance threshold is met. identifying a first possible PDF using a goodness of fit test; identifying a second possible PDF using a goodness of fit test; and choosing to use the first or second possible PDF based on a deviation from expected outcomes.

16. A method for determining a target stock level for an inventory of stock keeping units, the method comprising: running analytics and performing qualitative analysis to align a stock keeping unit ("SKU") to a "best-fit" planning model; statistically identify a probability distribution function ("PDF") for demand of the SKU; and assigning a deployment algorithm and lot sizing algorithm to deliver a target stock level ("TSL") to probably meet a service level at a desired cost objective.

17. A method for generating a sampling, comprising: determining a total population of an inventory of members; segmenting the total population into a plurality of segments; determining a sampling error for use in analysis; calculating a sample size for each of the plurality of segments; using a criterion measure for ranking members of each of the plurality of segments; generating a set of random numbers for each of the plurality of segments, wherein the set is equal to the sample size for the segment; and extracting from each of the plurality of segments the members whose rankings match the set of random numbers

18. A computer program stored on a computer readable medium for programming a computer to optimize target stock levels for an inventory of stock keeping units ("SKUs"), the computer program comprising: a code segment for developing a service strategy for assigning a service level for each of the plurality of segments; a code segment for analyzing the plurality of segments and the service levels for the plurality of segments for identifying at least one logistically distinct business ("LDB"); a code segment for assigning each of the plurality of segments to a "best-fit" planning model for indicating deployment, replenishment, forecasting and review characteristics for the segments; and a code segment for identifying a probability distribution function ("PDF") for estimating a demand process of each of the plurality of segments.

19. A computer program stored on a computer readable medium for programming a computer to meet a service level at a probable minimum cost, the computer program comprising: a code segment for identifying a probability distribution function ("PDF") of a demand process; a code segment for running a simulation to test the PDF against historical usage data to determine a service level; a code segment for adjusting a deployment algorithm based on results of the simulation; and a code segment for repeating the steps of identifying, running and adjusting until an acceptance threshold is met. 20 A computer program stored on a computer readable medium for programming a computer to determine a target stock level for an inventory of stock keeping units, the computer program comprising: a code segment for running analytics and performing qualitative analysis to align a stock keeping unit ("SKU") to a "best-fit" planning model; a code segment for statistically identify a probability distribution function ("PDF") for demand of the SKU; and a code segment for assigning a deployment algorithm and lot sizing algorithm to deliver a target stock level ("TSL") to probably meet a service level at a desired cost objective.

21. A computer program stored on a computer readable medium for programming a computer to generate a sampling, comprising: a code segment for determining a total population of an inventory of members; a code segment for segmenting the total population into a plurality of segments; a code segment for determining a sampling error for use in analysis; a code segment for calculating a sample size for each of the plurality of segments; a code segment for using a criterion measure for ranking members of each of the plurality of segments; a code segment for generating a set of random numbers for each of the plurality of segments, wherein the set is equal to the sample size for the segment; and a code segment for extracting from each of the plurality of segments the members whose rankings match the set of random numbers.