36 Association of Research Libraries Research Library Issues 298 — 2019 our data curators are able to better understand the specific needs of disciplinary research projects before they are funded. They can establish relationships with project personnel that will in coming years bear fruit in well-described, highly accessible, data submissions to the Ag Data Commons resulting from the funded work. Recent analysis by Smale et al. finds that supposed benefits of DMPs have not yet been supported by evidence.26 However, they suggest that placing DMP preparation into a researcher-centric context of education and sound program management may be useful. As this experimental service progresses we can determine if this is the case. Conclusion In summary, we have two primary conclusions: 1. It is important for a library like NAL to team up with universities and science organizations to explore new services and create talent pipelines. 2. It is productive to explore new services in the context of the entire data life cycle. These efforts allow us to gather the requirements for future programs that may or may not need to be housed in the library. In closing, let us consider the benefit to the library of developing these services here rather than being embedded, for example, in an IT services organization or a scientific department. Developing data science services in a research library—in the context of other data management services—provides an excellent way to attract and develop technical talent to the field of library and information science. It may even be the case that data science can help increase diversity in the library pipeline. People of many backgrounds, genders, and races may be excited by a career in innovative science and technology and find satisfaction in practicing those skills in a mission- driven service context rather than profit-driven organization. Finally, given the uncertain future of traditional scholarly publishing, the role