Celebi, HeidiRetail shelf availability research has been limited by the inability to measure stockouts. Not being able to fully capture stockout occurrences has led to studying either the effects of stockouts or their antecedents. It has also led to using various fundamentally different stockout attributes as measures across studies. The relationship between stockout attributes is not clear, making it difficult to have a consensus on either the drivers or the impact of stockouts. This thesis considers both antecedents and effects of stockouts by incorporating actual stockout events under two different risk pooling methods. The first set of models simulate stockout-based customer switching (the inventory effect) to study pooling by substitution for a retailer setting service level goals for two products. The second set of models study pooling by postponement, termed “instore logistics postponement,” using archival data from a new shelf sensor technology that captures actual stockout events. An extension to the second part of this study examines the nonlinear relationship between stockout attributes. Both parts of the dissertation contribute to the stockout literature in different ways. The simulation work contributes towards reconciling opposing views on the performance effect of risk pooling through substitution, also showing how different performance measures may accentuate or mask the impact of stockouts. The shelf technology work contributes to logistics postponement by studying how a two-tier inventory within the store may affect stockouts along more than one stockout attribute, and whether less frequent but longer stockouts are linked to better performance than shorter but more frequent stockouts.enDissertationBusiness administrationOperations researchPostponementProduct AvailabilitySubstitution