2016 Charleston Library Conference

Demand Driven Acquisitions (DDA) programs have become an integral part of academic libraries' collecting strategies. While DDA programs provide an effective way to build a just in time collection, it can be difficult to anticipate how many titles will be triggered for purchase and what the financial impact will be. This presentation will describe a project to build a predictive model to flag DDA titles that are likely to be triggered for purchase within the first year of being added to the catalog. By implementing a predictive model, collections and acquisitions departments can better plan the yearly DDA budget. In addition, titles with a high probability of being triggered for purchase can be purchased if they become ineligible for DDA. We will discuss how we combined text analytics and structured data as inputs to the model using a combination of SAS and Python. In addition to the benefits of implementing a predictive model, we will also discuss the drawbacks and limitations involved.