טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentAssaf Avrahami
SubjectThe Value of Perfect and Imperfect Information in a Multi-
Location Inventory System
DepartmentDepartment of Industrial Engineering and Management
Supervisors Professor Herer Yale
Full Professor Shtub Avraham
Full Thesis textFull thesis text - English Version


Abstract

The fact that information has value in the management of supply chains is well accepted. Our work focuses on two aspects of information: quality and quantity.

In examining the quality of information, we are concerned with discrepancies known as inventory errors, which are the result of events that can be classified as one of the following three types: Shrinkage, Misplacement and Wrong Scanning. These events lead to discrepancies between inventory reflected in the IT system records (the computer records) and available inventory for sale. The literature describes both the extent to which these types of discrepancies are commonplace and the degree to which they cause increased lost sales and lower profits. Nevertheless, these discrepancies and/or their associated negative effects can be eliminated if more information is available.

Our work explores four different models that operationally take these errors into consideration in four different ways: the no information scenario, the static informed scenario, the informed scenario and the full information scenario. The objective of our work is to explore and quantify the value of the additional information in all three scenarios compared to the no information scenario. We develop an analytical model for each scenario, where the objective in all four scenarios is to minimize the average expected cost per period. We find the optimal order-up-to level for each period. In addition, we find the optimal frequency for performing inventory counts for the no information scenario, the static informed scenario and the informed scenario. There is, of course, no need to conduct an inventory count under the full information scenario.

Our findings indicate that additional information greatly influences optimal policies and, if exploited wisely, can save a great deal of money. Moreover, we find that certain problem parameters do not affect the value of information (e.g., the expected value of the demand), whereas other parameters influence it tremendously (e.g., the expected value of the mean shrinkage).

For our examination of the quantity of information, we focused on distribution systems that are based on a network of retailers. We sought to explore and quantify the value of additional information in these systems. In particular, we explore the value of our ability to review the state of the system more frequently and in this way to allow the partial aggregation of the retailers.

Intuitively, this research seeks to implement risk pooling through the partial aggregation of the retailers. From an operational point of view it would be best to fully aggregate the retailers, i.e., effectively combining the retailers into a single retailer. In our case, this aggregation is not possible since each retailer is owned by independent entities with no connection between them. Our model allows partial risk pooling through “virtual” risk pooling, i.e. holding back some inventory until some of the demand is revealed.

In our study we develop a natural formulation for the problem and later on a working formulation for our model. We prove convexity of our model and find an optimal solution for the problem. We developed an algorithm to numerically solve the problem and programmed the algorithm using Matlab. We performed a numerical study, numerical analysis and large-scale field study. We report on the savings that the additional information enabled (i.e., the value of the additional information) and discuss in detail what we learned both about the original system and the information rich system.