In today’s environment of powerful personal computers and software, developing new specialized software can be much easier. Accessing files and creating routine tabulations that once took teams of individuals months to produce can often be produced by a very good researcher or programmer in a few weeks. New statistical methods such as Fellegi-Holt edit/imputation, Fellegi-Sunter record linkage, confidentiality, and statistical methods such as bootstrap, multiple imputation, Expectation- Maximization (EM), and Gibbs sampling require both detailed knowledge of theory and the development of new computational algorithms. In the advanced statistical situations, there is much greater need for programming skills that are consistent with the analyses and theory. If we consider those successful software projects at different statistical agencies where these methods have been applied, it is likely that there is a key common characteristic. The common characteristic for developing specialized software is the project team or individual who understands the theory, the required analyses, and can do the programming.