Federal administrative data is a low-cost and low-burden data source for evidence-based policy making. By linking information from different surveys, or over time, researchers can achieve the sample size and variation needed for advanced econometric methods. However, the personally identifying information (PII) needed to link information means that these data re not available to the public. One solution is to provide technical specifications to the requisite agency(s) to execute the research. This paper outlines the process and pitfalls of drafting specifications for an implementing party who knows more about the data than you do. Drawing on experience from working with the U.S. Census Bureau and knowledge gained from related literatures, such as open-source coding, this paper recommends the depth of description, order of data manipulation and analysis, and requested output to make these collaborative projects successful. A federal administrative data project proposal template is offered. The paper also advises on information that federal agencies can supply to facilitate the use of these important data sources.