Supply Chain - Inventory Optimization - Case Study Example

Supply Chain: Inventory Optimization

Introduction

It is the wish of every supply chain manager to know the exact nature of the future events. Such knowledge will be vital in ensuring that the manager takes actions accordingly or plan for actions when sufficient time is at hand and, therefore be able to implement the plan. However, the effectiveness of plans depends on the level of accuracy of the knowledge of future events known to the manager. Managers thereby rely on forecasts to make management decisions. Such estimates are, however, mostly affected by errors making inventory optimization one of the key challenges in supply chain management (Chen, Li & Shady, 2010). The above situation leads us to the question, "is it possible through improving inventory management to find the adequate quantity of stocks that will fulfill demands, avoid overstocks, and yet lower the cost of capital?" This is the question that this study is striving to answer. To answer the question above; we are going to consider a case scenario of Intel's Channel Supply Demand Operations, a business entity that has launched an initiative to improve its supply chain performance.

Case Study: Identification of Issues and Problems

Intel has been making efforts to improve its inventory planning processes. The improvement efforts have been through two ways, (a) developing an integrated sales, inventory, and operations process (SIOP) and (b) improving the decision-support technologies by integrating multi-echelon inventory optimization with its supply chain planning solvers. The development of sales, inventory, and operations process (SIOP) requires the firm to define its strategic initiatives, business metrics, and measurements (Kampen, Donk & Zee, 2010). To achieve this, the SIOP process must be embedded in the organization's existing structure by determining who is responsible for what in the initiation of demand planning, estimating sales and the team capacity. This will ensure that there is an able team to execute and streamline the expectations and implementation of the processes throughout the different levels of the organization.

However, during the implementation of the SIOP process, the management encountered three forecasting challenges that had significant impacts on the optimization models including forecast bias, heterogeneity of errors, and nonparametric residuals. To achieve a proper inventory planning, the supply chain managers had to address the three identified forecasting problems (Manary, Willems & Shihata, 2009). The demand forecasts, which are crucial to supply chain management, form an essential input to the inventory planning processes. Such errors within the demand forecasts undermine the reliability of the inventory planning appraisal. The minimization of demand forecasting errors is, therefore, important in achieving lower inventories, reduced stock-outs, improved customer services, smoother production plans and reduced operation costs.

To improve inventory forecasting accuracy, the company employed a program that would procedurally modify the original data being fed into the firm's forecast system. The program consisted of statistical methods for testing heterogeneity of independent sample variances, an application that extracts random samples of variance estimates for an individual product among the many service levels. In cases where no significant variation is detected in the variability estimates, a weighted proxy of variability was applied. The application of the weighted proxy of variability is carried out through the root mean of the combined modification procedure at the sampled service levels (Kampen, Donk & Zee, 2010).

The adjustments were applied with the assumption that forecast bias was stationary, and service levels were predetermined. Through the application of the kernel smoothing technique, Intel was able to treat stock-keeping units (SKUs) that possessed considerable differences in variability. The kernel smoothing technique generates a single estimate of variability by considering both heterogeneity and nonparametric error distributions (Kampen, Donk & Zee, 2010). In this way, Intel managed to overcome the presence of forecast bias by making adjustments to the estimates of the standard deviation of forecast error and thus achieving the desired safety stock and service-level targets. Despite the challenges in reaching the optimal inventory accuracy, Intel managed to make a reduction in the safety stock levels by averagely 15 percent without altering the raw demand signal.

The modification procedure also provided the requisite improvement for the implementation of the multi-echelon inventory optimization (MEIO) models and the embedment of the same into Intel's processor division (Manary & Willems, 2008). Given the function of SIOP of enhancing coordination by facilitating collaborative decision making across the enterprise with the aim of aligning production with sales, it created a good condition for the implementation of MEIO. With the large number of production and different tiers or echelons of the enterprise's distribution network, management of inventory is often a daunting task for Intel. Multi-echelon inventory optimization (MEIO, therefore, ensures that the safety stock buffers are right-sized across the entire supply chain and considers the interrelationships between stages, as well as variables that cause periodic excess inventory, such as continued demand uncertainty, lead times, and supply volatility.

The application of multi-echelon inventory optimization (MEIO) was aimed at improving the accuracy of set inventory targets in the Intel's embedded device group. The accuracy in setting inventory targets was expected to reduce inventory costs simultaneously and raise customer-service levels. Historically, Intel's integrated device group's inventory target setting was management determined and was applied uniformly across product families. However, it was difficult to achieve and maintain the inventory targets at the individual product level. The challenge called for the employment of a multi-echelon inventory optimization (MEIO) model to set inventory targets in achieving a better alignment of inventory resources and improvement in customer service levels (Kampen, Donk & Zee, 2010).

The management, however, encountered a hurdle in implementing the multi-echelon inventory optimization (MEIO) model's initial recommendations due to biases in the sales forecast data. The challenge was faced because the modified raw sales forecast data could not be a useful option as sales and marketing data was loaded into the manufacturing resource planning (MRP) system before it was received by the planning organization (Manary & Willems, 2008). The raw sales and marketing data was thereby out of the control of the executive, and they were, therefore, unable to make adjustments. What this meant is that the bias was already in the system and resultantly in the average forecast demand. To resolve the error already created, the planners had only one option which was to adjust the inventory targets. This idea enabled the planning organization to overcome the inventory optimization problem in Intel's embedded devices group through the production of appropriate inventory targets even in the presence of forecast bias. Therefore, Intel's case scenario is a proof that computation coupled with necessary adjustments can be used to improve inventory management for organizations. This is achievable as organizations can find the quantity of stocks that fulfill demands, avoid overstocks, and yet lower the cost of capital.

Case Analysis and Evaluation

Determination of the safety stock within the supply chain is the primary challenge in inventory management (Chen, Li & Shady, 2010). Safety stocks, as shown in the discussions above are crucial in mitigating the effect of variability in demand and supply so as to uphold the desired level of customer service. Safety stocks are usually unnecessarily high in many supply chains and thus offer an excellent opportunity for cost reductions by reducing the excess stocks (Manary, Willems & Shihata, 2009). The case above is an example of how modern applications such as the sales, inventory, and operations process (SIOP) and multi-echelon inventory optimization (MEIO) models can be useful in inventory planning process. It is important to understand how the adoption of the two concepts (SIOP and MEIO) can smoothen the supply chain management decisions. In order to analyze and understand the case scenario above, we will need to understand the following: (a) how the inventory planning process works (An Overview of Inventory Planning), (b) how the two supply chain management support models, SIOP and MEIO, functions, and (c) the dilemma that companies, with reference to Intel, face in inventory planning process.

• An Overview of Inventory Planning

In inventory planning process, the key optimal inventory inputs, inventory safety stock, and re-ordering levels are valuable parameters imperative in the control of the two significant factors in the supply chain, the ability to uphold a favorable service level and the amount of inventory to be held. Decisions on the level of stock safety command for a trade-off between inventory costs and stock-out costs with considerations given to the uncertainties (Wieland et al., 2012). However, the balance between inventory costs versus stock-outs costs is very business dependent. Therefore, instead of giving considerations to the two costs directly, a classical notion of service level is introduced. The service level is an expression of the probability that a certain level of safety stock will not result into a stock-out. Naturally, service levels increase with increases in safety stocks. There is thus zero likelihood of encountering stock-out (meaning that service levels will be inclining to 100 percent) which is achieved as the safety stocks get larger. This relationship is as shown in the figure below:

(Kampen, Donk & Zee, 2010)

From the above illustration, it is notable that the required optimal inventory levels that are necessary for guaranteeing the desirable service levels change as the demand and supply changes. The ability of the supply chain managers to service demand is directly dependent on the safety stock (Weele, 2010). This is so due to the inherent variability in the demand and supply streams experienced at any supply chain mode. The importance of a proper inventory planning process is to identify the service levels and differentiate the products that call for higher service levels from those that do not. Such a take is imperative in upholding user-defined service levels that guarantee a sought-after fill-rate to realize the demand. A proper inventory planning process also reviews such service levels and considers changes to the safety stock recommendations in response to an adjustment of demand and supply (Chen, Li & Shady, 2010). A quick synopsis of the optimal inventory planning process, as discussed above shows that the inputs into the inventory optimization are desired service level, demand, supply and supply lead time.

• Sales, Inventory, and Operations Process (SIOP)

The function of SIOP is to enhance coordination by facilitating collaborative decision making across the enterprise with the aim of aligning production with sales. The model operates by identifying anticipated changes, making necessary adjustments to mitigate its associated risks and thereby seizing new opportunities. A cross-functional process of SIOP involves customer and product segmentation, product management review, demand review, inventory analysis, supply review, and executive decision making (Chen, Li & Shady, 2010).

Segmentation of customers is the realization that the clients do not have similar needs. On the other hand, product segmentation means that all the goods are not equal as some are easier to produce, more strategic, and easily substituted by an alternative or customer specific. Ranking of clients and products regarding priority is, therefore, necessary as there is no one-size-fits-all in the supply chain (Weele, 2010). Segmentation of customers and products is important in clarifying choices (what is more important) to have a more efficient supply chain. Product management review enables the management to identify changes in the product and how such changes affect the market performance of the product and also other changes that the organization can make to improve the quality of a product. Therefore, inventory, demand, and supply are closely related because the knowledge of demand will be a reflection of the amount of stock that should be held (Wieland et al., 2012). The amount of inventory will, in turn, dictate the ability of the company to supply and satisfy its customers. Finally, executive decision making involves reviewing the whole process and adjusting where necessary to have an accurate estimate of the inventory required.

• Multi-Echelon Inventory Optimization (MEIO) Models

Multi-echelon inventory optimization (MEIO) model is a tool for calculating and setting supply chain safety targets. This method uses sophisticated algorithms to calculate the targets in a multi-echelon way. The targeted safety stocks are significant in achieving the desired customer service levels while avoiding the additional high risk safety stocks. In this way, the amount and cost of inventory that is necessary to attain the targeted service levels can be reduced substantially. It is imperative to understand the difference between a multi-echelon inventory optimization (MEIO) and the traditional ways of managing inventory (single-stage inventory optimization) to appreciate how the concept of MEIO leads to better results.

A single-stage inventory optimization is different from the multi-echelon inventory optimization because it optimizes safety stocks locally as a single location of the supply chain as opposed to MEIO that gives a holistic view of the supply chain (Kampen, Donk & Zee, 2010). The single-stage inventory optimization's tendency to have a localized view of the supply chain leads to poor coordination of the entire supply vision. A multi-echelon inventory optimization is thereby necessary considering that the nature of supply chains heavily rely on outsourced activities.

The strength of multi-echelon inventory optimization (MEIO) is in its ability to provide the right-sized safety stocks for the entire supply chain. This is achievable by the model's consideration of all the interdependencies between stages and the different variables responsible for the chronic excess inventory such as demand uncertainty, supply volatility and long lead times. The data that is necessary for a multi-echelon inventory optimization project include product location master (PLM), bill of materials (BOM) and demand profile (Kampen, Donk & Zee, 2010). The data on the bill of materials describes what is involved in the production of the materials while product location master provides the link between the manufacturing plants, product, and suppliers. On the other hand, demand profile provides a means of exploring the customer demand for the product. Hence, an understanding of the demand profile, product location, and bill of materials helps in the better estimation of target safety stock determination and enables the management to achieve optimal service levels while minimizing supply chain costs.

• Intel's Supply Chain Dilemma

Continued vitality characterizes the global economy in which Intel operates in as a result of increased competition and unpredictable customer demand (Kampen, Donk & Zee, 2010). Amidst such economic turbulences, the focus of the corporation's executives has shifted from revenue growth to profitability growth thereby giving the performance of the supply chains a lot of attention. Intrinsically, the primary aim of the businesses is how to devise ways and means of responding to the customer’s demand, uncertainties, and the harsh competition. This would involve overcoming the challenge of achieving inventory optimization, while upholding higher customer service levels and reduced variable costs.

Two closely related processes are imperative to the company's achievement of optimal inventory levels that are necessary for meeting expected service levels, the replenishment (or re-ordering) process, and the inventory planning process. The two processes are closely related since inventory planning process provides the decision parameter upon which replenishment will be based so as to achieve optimal levels (Chen, Li & Shady, 2010). Critical to the understanding of the relationship between the two processes is the realization that inventory stock depends primarily on two factors; demand (the amount of items that are expected to be bought or consumed) and lead time (which is the delay between re-order decision and renewed availability). However, the two factors above are subject to uncertainties which include demand variations, the ever-evolving customer behaviors that occur in rather unpredictable ways, and the lead time changes such as the unplanned difficulties that suppliers and transporters may face (Wieland et al., 2012).

If future demands were perfectly known and demand entirely reliable, then there would be no uncertainty on the reorder point, that is, the amount of stock that calls for a reorder. The situation should be as articulated above since the reorder point would be equal to the total forecasted demand during the lead time (Weele, 2010). In practice, such a scenario cannot be a reality because of the prevailing economic uncertainties. This is despite the sophisticated modern day system applications that have been designed to initiate and streamline the operations and inventory management. Nonetheless, the economic unpredictability has led to forecast biases, providing a chance that the future demand forecasts (the forecasted value) might be greater or lower than the actual demand. It is thus important to consider two elements, forecasting accuracy and bias, critical in evaluating forecasting performance (Chen, Li & Shady, 2010). The above two factors lead to the question, "is it possible through improving inventory management to find the adequate quantity of stocks that will fulfill demands, avoid overstocks, and yet lower the cost of capital?"

Recommendation

When forecasting, the accuracy of forecasts can only be the degree of closeness of the stated quantity to the actual quantity (true value). This is because there is no mechanism for measuring the actual value at the time the forecasts are made since they are just statements regarding the future (Kampen, Donk & Zee, 2010). It is, therefore, important to note that the only way that we can achieve that nearness to the actual quantity (the accuracy of the forecast) for inventory optimization and demand planning purposes is to adopt a statistical viewpoint. The lack of future data necessary in achieving actual values leaves us with only one option of estimating the future demand, the time series forecasting method. This approach utilizes historical demand to make a forecast for the future market. The approach is based on the assumption that past demand history is a good predictor of future demand. However, the method is appropriate when the fundamental demand trends do not experience a significant variances from season to season.

The term forecast accuracy in this case, therefore, is a computed estimate of the quantity of the expected quality of predictions that is strictly measured against the data that we do not have yet, as opposed to the quality of a physical measurement of some kind. This makes forecasts prone to errors as shown in the case study scenario and the analysis. Forecast errors are thus common when making estimates and result from many sources one of them being systematic. On the other hand, a prediction error denotes the difference between the actual demand and the forecast demand (Manary & Willems, 2008). It is always stated as an absolute value or as a percentage and can be an over-forecast or an under-forecast. An over-forecast means that the forecasted amount is more than the actual required amount while an under-forecast indicates that the forecasted value is less than the actual.

To achieve accuracy in forecasting demand in order to improve inventory management and find the adequate quantity of stocks that will fulfill demands, avoid overstocks, and yet lower the cost of capital, a manager (estimator) is required to collect the relevant historical data, evaluate and analyze the data appropriately, make rough forecasts based on the collected data and use statistical methods to correct the errors that might have cropped in. The last step, modification aimed at correcting errors, is vital to the forecast process because it increases the confidence of the stakeholders who want to know if they can trust the forecasts and gain recommendations on how to apply them to improve their strategic planning process.

Correcting of error can be done by the use of statistical accuracy metrics of measurements of forecasts accuracies, which include the mean absolute error (MAE) and mean absolute percentage error (MAPE) among other techniques (Chen, Li & Shady, 2010). These measurements of error have a critical role in tracking forecast accuracy, monitoring for exceptions, and benchmarking the forecasting process. Increasing forecast accuracy is, therefore, a competitive necessity that is crucial to the achievement of top-line growth as well as bottom-line profitability. It is an essential step to reducing inventory levels and its associated carrying costs. It is also significant in managing scrap and ensuring improved customer service levels by having the right product in the right place and at the right time.

Conclusion

As seen in the discussions above and the Intel company scenario, supply chain models such as sales, inventory, and operations process (SIOP) and multi-echelon inventory optimization (MEIO) can improve inventory planning processes. However, these models are prone to biases while the decisions that they support can also be subjected to forecasting errors. These biases can be eliminated by making a necessary statistical adjustment to come up with a more accurate supply chain forecasts. In conclusion, it is possible through improving inventory management to find the adequate quantity of stocks that would fulfill demands, avoid overstocks, and yet lower the cost of capital.