In this thesis, three procurement policies (Backup Policy, Blend Policy, and Partial Order Policy) are considered and evaluated using a Discrete Time Markov Decision Process. The following factors were analyzed: retailer's demand distribution, holding cost, purchasing cost and supplier's characteristics like the risk of disruption, and the repairing probability. Of these three policies, Partial Order Policy gives the decision maker the most flexible selection on suppliers and order amounts. Thus, it is observed that it always obtains the largest net revenue among the three policies. However, in the examination of this policy with variables of purchasing cost and risk probability, the decision maker does not split the order every time. The order is split only when the retailer's inventory level is lower than the Threshold Inventory Level. In addition, when the inventory is over Target Inventory Level, the decision maker does not place an order with either supplier. Experimentation with factors such as α, c2 and h, the Threshold Inventory Level and the Target Inventory Level is significantly impacted. Especially, when h is very expensive and c2 is getting close to c1, the best policy would be changed by placing orders with the reliable supplier until the retailer has enough inventory. The Discrete Time Markov Decision Model in this thesis offers a competitive and flexible way to compute the retailer's ordering policy in each state to obtain the best procurement policy.