Date of Graduation


Document Type


Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level



Industrial Engineering


Nebil Buyurgan

Committee Member

Edward A. Pohl

Second Committee Member

M. Alp Ertem

Third Committee Member

Shengfan Zhang


Applied sciences, Cycle count, Error, Fuzzy, Inventory record accuracy, Markov decision process, Multi-objective


This dissertation performs a methodical analysis to understand the behavior of inventory record inaccuracy (IRI) when it is influenced by demand, supply and lead time uncertainty in both online and offline retail environment separately. Additionally, this study identifies the susceptibility of the inventory systems towards IRI due to conventional perfect data visibility assumptions. Two different alternatives for such methods are presented and analyzed; the IRI resistance and the error control methods. The discussed methods effectively countered various aspects of IRI; the IRI resistance method performs better on stock-out and lost sales, whereas error control method keeps lower inventory. Furthermore, this research also investigates the value of using a secondary source of information (automated data capturing) along with traditional inventory record keeping methods to control the effects of IRI. To understand the combined behavior of the pooled data sources an infinite horizon discounted Markov decision process (MDP) is generated and optimized. Moreover, the traditional cost based reward structure is abandoned to put more emphasis on the effects of IRI. Instead a new measure is developed as inventory performance by combining four key performance metrics; lost sales, amount of correction, fill rate and amount of inventory counted. These key metrics are united under a unitless platform using fuzzy logic and combined through additive methods. The inventory model is then analyzed to understand the optimal policy structure, which is proven to be of a control limit type.