Inventory Management and Re-Order Point Analysis at the Client S Component Re-Build Center

Inventory Management and Re-Order Point Analysis at The Client’s Component Re-Build Center

Prepared by:

Kelley Bessette
Jeff Litchfield
John Johnson

Matt Riskin

April 2002

Introduction

The Client is one of the world’s largest dealers of heavy machinery products. One subdivision of The Client is the Component Rebuild Center (CRC), which is located in Edmonton. At the CRC, used parts are refurbished and sold to existing clients. The Client is interested in accurately assessing the number of refinished parts needed throughout their network to satisfy demand and allocate as little capital as possible to dormant inventory. Customers are able to sell their used parts back to the CRC when they buy a remanufactured part from the CRC, which makes it a closed loop system. The production manager of the CRC was concerned that they were holding too much finished goods inventory.

Terminology

Throughout the duration of this project, the following terminology played a major role in the understanding of the problem and solution.

·  Branch – Any node in the network.

·  Demand Center – Any branch in the network with demand over a user defined critical number of parts per year.

·  Distribution Center – Services the branches by carrying inventory for those which do not qualify as demand centers. It is of note that sales occur here as well, thus inventory carried here serves both its own sales, and those for the branches for which it carries goods.

·  Core – Any part in the system.

·  Core Bank – Cores waiting to be refurbished at the CRC.

·  OR Sales – This is a lost sale due to not carrying adequate inventory. In this case, a customer goes directly to Caterpillar or another competitor to purchase a part.

Problem Definition

The Client sells, services and finances the full line of Caterpillar and complementary equipment throughout British Columbia, Alberta, Yukon, and the Northwest Territories. Below is a map of the area that The Client services.

When a The Client customer needs an existing part replaced, they have the option to make use of a used part exchange program. This entails selling back the used core in return for one of The Client’s refurbished parts. The old core will be returned to the CRC where it is remanufactured and made available for sale. This results in a closed loop inventory system, as is shown below.

Original Inventory Flow

The CRC services five major distribution centers, and over forty-five satellite branches throughout this area. Inventory decisions must be made for over nine hundred parts, making inventory management at this center a very complex process. The three particular areas of concern that the management at the CRC asked us to address for each individual part were:

1.  How many total cores to have in the system

2.  Where in the network finished inventory should be stored

3.  The re-order points to maintain at the individual branches and distribution centers

At the time we were presented with this project, The Client did not have a tool to aid them in making these decisions. The Client’s project team wanted a tool that would complement the decision maker’s expertise, and assist them in their daily production decisions. This tool had to be applicable to all nine hundred parts, and had to be developed in Microsoft Excel. A large portion of the Component Rebuild Center’s inventory is composed of smaller components of major machinery. The life cycle of these components is highly variable and makes the demand for these parts irregular and unpredictable.

Methodology

After initial meetings with the The Client project team and clarification of what our project goals were, it was decided a simulation model would be developed. Simulation is an appropriate method in ambiguous and unknown situations, especially when there are many complexities and conflicting variables. Due to the irregular and unpredictable nature of sales, and the lack of historical data, we determined that forecasting would have been very inaccurate. When simulation is used, the timing of demand occurrence is irrelevant. We wanted to mimic the overall demand and observe how a particular set of parameters would perform. For this reason, we decided to build a simulation tool that enables The Client to retrieve the information that they originally requested.

The downside associated with using simulation for this purpose is that generally simulation is more descriptive and should be used cautiously as a prescriptive tool. The model attempts to develop recommendations by using an interative process and multiple runs of the simulation to analysis various possible decisions. The performance of these decisions is measured on criteria that the The Client management team has deemed most important to the operational efficiency of the company. Although simulation is not an optimizing tool, the results provided will offer a clear picture of how the system functions and the recommendations can be applied with less cognitive adjustments than previously required.

Assumptions of the Model

In order to build a tool that would be flexible enough to be applied to all 900 parts, a number of assumptions had to be made. The first assumption is that all finished goods would be routed through the main distribution center located in Edmonton, referred to in The Client’s system as number 45, as is shown below.

Modified Inventory Flow

Another assumption of the model is that parts have an infinite life. In reality, a core will reach a point where refurbishing is no longer possible. At this point, the core is scrapped out and a new core is introduced into the system to take its place. After consulting with the management team, it was decided that for the purposes of this model, this aspect of the system would not be considered. An additional assumption that the model makes is that the most favorable re-order point is obtained when the cost of holding an additional unit becomes greater than the revenue generated by that unit. Furthermore, we assume that the cost associated with loss of goodwill is simply equal to the revenue lost from the lost sale. The final major assumption of the model is that future sales will resemble past sales.

User Inputs

A portion of this project was spent determining which parameters should be dynamic. Along with the management team at The Client, the inputs themselves, and the way they were to be incorporated in the model were determined. The model sets these inputs to default values, but can be manipulated if the user chooses to do so.

Ø  Main Distribution Center

Currently the main distribution center is located in Edmonton. This was added as a user input so that if the location of the main distribution center were to ever move from 45, the model could be adapted accordingly.

Ø  Number of Cores in the system at the beginning of the simulation

There are two ways to use this input. The default setting involves initializing the model with an arbitrarily high number. This is to ensure that there are enough cores in the system to satisfy all demand. Throughout the multiple runs of the simulation, the fluctuations in the number of cores that are actually required are tracked and an average of this is taken. For the final run, when the ideal re-order points have been established for each branch, the number of cores is scaled back to this average number of required cores.

This input can also be used to perform sensitivity analysis. The user is able to specify the number of cores in the system. In this case, the number of cores does not fluctuate at all, enabling the user to observe the effect of keeping a static number of cores in the system.

Ø  Core process time

This is the time it takes to remanufacture a core. This value can be inputted as a static value, or the user can select one of the following distributions to be applied to the model:

§  Normal
§  Uniform
§  Deterministic / §  Triangular
§  Poisson
§  Exponential

Ø  Process Capacity

This input allows the client to put an upper bound on the number of parts that can be processed for a particular part at any one time.

Ø  Costing Information

The end user inputs all costing information. The relevant inventory costing information for this model is holding cost per year, investment cost per year and lost contribution per unit. This is input as a yearly cost and translated into a daily cost for the purposes of the simulation.

Ø  Number of Cycles in the Simulation

The number of cycles is essentially the number of days the model will simulate. The higher this number is set, the more reliable the results will be. It should be noted that there is an upper bound on this input of 2000 cycles.

Ø  Demand Multiplier

Because the simulated demand is based on past sales, the demand figures will be consistent with them. In cases where the client is anticipating a change in sales levels, the user is able to use the demand multiplier to magnify demand.

Ø  Sales Criterion

During the data extraction phase of the model, average annual sales for a core is computed for each branch in the network. If this number does not equal or exceed this sales criterion, then the part will not be housed at the branch and will instead be stored at the local distribution center.

Ø  Used Part Return Time

We have implemented a discrete distribution for the return time that the user is able to change, as they feel necessary. This can be used to evaluate the impact of implementing a formal core return policy.

Ø  Network Information

The network information lists every branch and includes which demand center services it. The user is able to update any changes in the network information directly in the model.

Ø  Process time

This is the time, in days that it takes to refurbish the part being run. This is a static figure.

Simulation Process

Based on the assumptions stated above, the model is a simplification of reality. It possesses the aspects which are relevant for the purpose of our analysis. After the model generates this information, the simulation portion of the model is activated. The simulation itself accounts for a number of possible occurrences, including:

1.  Simulation of sales at demand and distribution centers

Upon receiving demand figures for fifteen parts, we analyzed the data to determine if there was a distribution that could be used in the simulation of demand for all parts. We found that past sales for the parts that were analyzed were highly sporadic, and showed substantial variation in their magnitude.

After researching these types of demand profiles it was determined that the most applicable simulation method would be to use a variation of Croston's Intermittent Method. This method simulates demand in two steps. First, the probability of a sale occurring is calculated. An assigned distribution is then used to determine the magnitude of a given sale.

Past sales information for the part that is being analyzed is retrieved by the user from a company database. The model uses those figures to generate the probability of a sale and its magnitude. For the purposes of simulating the magnitude of demand, it was determined that the Poisson distribution would suit the demand profile most accurately.

Within our model, a given sale can be incurred at either a branch which qualifies as a demand center or a Distribution Center.

2.  Lost sales

The second component of the simulation model is to track the number of lost sales in the network. The model recognizes a lost sale whenever demand is occurred at a location, and the inventory to meet it is not available. When this occurs, the value of the lost sale is added to the lost contribution cost. There is no cost associated with loss of good will taken into consideration.

3.  Return time for used

When demand is incurred in the simulation model, an associated used core is sent back to the CRC for refurbishing. In reality, there is a delay associated with this return time. In the model, the core is assigned a return time using a discrete distribution, which is defined by the user. All returning cores are assigned a return time; and this delay is incorporated into the model. At this point, the core arrives at the CRC, where it will be processed.

4.  Process time for cores at the CRC

Once a core has arrived back at the CRC it is put through the refurbishing process. Although this process does have a certain degree of variation in the processing time, this is to be a set time that would be input prior to running the model.

5.  Time to Distribute Cores

Once a core is remanufactured, it is put into finished goods inventory. All finished goods are routed to the main distribution center in Edmonton prior to being shipped out to the appropriate demand centers in the network. For the purposes of the model, distribution from here to each demand center takes a fixed amount of time.

Financial information for simulation

The financial outputs of the model are a direct function of the cost inputs. The model tracks the total number of days that finished inventory exists at the distribution center in Edmonton, en-route to various demand centers and distribution centers, and any inventory on hand at these centers. The following is an explanation of how the financial information is incorporated in our model.

Annual investment cost

The annual investment cost is the cost of having a core in the system. Because this is a closed loop system, The Client incurs this investment cost even when the core is in the customer’s possession, waiting to be returned. Thus, this cost accumulates throughout the simulation regardless of the state the core is in.

Annual holding cost

The annual holding cost is the annualized cost of processing a core. It is the cost of investing CRC time and resources to refurbish the cores. This cost begins to accumulate as soon as the core has been refurbished and does so until a customer purchases that core.