In warranty inventory management, customers return allegedly malfunctioning units to a company for replacement or credit. Useful units may be recovered through testing and/or remanufacturing processes; the company can use these recovered units to fulfill future warranty requests. The company also has the option of purchasing new units from a production line. In high-volume situations, warranty inventory management involves many complexities such as stochastic demand rates, probabilistic requests for credit instead of replacement, probabilistic repairs, multiple sources of supply originating from both the stochastic reverse channel and the company's purchasing decisions, and, in some cases, batching of remanufacturing. First, we formulate several related warranty inventory system models that aim to minimize inventory levels while meeting service level constraints. These models study periodic, single-location, inventory systems that are dedicated to warranty returns and include the following complexities: random warranty claims, random requests for replacement or credit, three sources of supply (testing, remanufacturing, and new product), random flows of returned products into testing and remanufacturing, random yields from testing and remanufacturing, different lead times for each resupply process, remanufacturing lead time variability, and random batching of remanufacturing. Using well-developed heuristics, these models produce results that provide near-optimal inventory-control policies in this complex environment and demonstrate the payoffs that result from reducing production lead times and batching in remanufacturing. Next, we consider several related models that incorporate the intricate cost structures that are often present in warranty inventory systems. We formulate two-stage models in which the company first makes a strategic decision on testing and remanufacturing capacities; then, in the second stage, the company aims to minimize inventory costs when faced with various levels of visibility into pipeline inventory in the reverse channel. "The Curse of Dimensionality" prohibits us from solving for optimal policies in most practical cases. Thus, we develop heuristic dynamic programs that allow for tractable models while incorporating information gained from the reverse channel visibility; we call this latter concept Advance Supply Information. Supplier decision models. Learning curves are a well-studied phenomenon in both the theoretical and empirical disciplines. As it gains experience in manufacturing a product, a factory/supplier is able to reduce that product's production costs. Buyers can choose their factory locations based upon the rate of learning that occurs at the factory. Many times, the lowest cost factory today becomes the high cost factory tomorrow due to slow learning. In order to explore these relationships, we have created several related models that solve for the optimal purchasing decisions under various supply chain configurations. We also evaluate the cost of ignoring the learning curve. First, we explore a model with deterministic parameters and then expand the model to include a stochastic learning portion.