Initial deployment decisions are critical to humanitarian organizations as they pre‐position resources and prepare for predictable disasters, such as hurricanes, floods, and wildfires. Deploying too few resources results in unmet demand, bad publicity, and unhappy donors. However, deploying a large number of resources can also result in bad publicity and unhappy donors due to high costs and wasted resources when disasters do not occur as expected and deployed items are not managed efficiently. We use a stochastic optimization model to investigate the initial deployment decision and resulting costs. The stochastic model forces the initial deployment decision to be made after the disaster warning, but before the disaster occurs and true demand is realized. We find that a full optimization model (incorporating returns and disposal into the model) will increase the initial deployment level from the level obtained without incorporating returns and disposal (referred to as a partial optimization model). This allows more beneficiaries to be served when using a full optimization model. We also find that fixed return costs do not have a large impact on initial deployment levels, but can result in a more consolidated pre‐positioning strategy. Using the results of our models, we present several other interesting managerial insights that may be useful for humanitarian organizations in pre‐positioning inventory for predictable disasters.