Policy making, while not simple, can be pursued in a straightforward way, with traditional phases of data gathering and analysis. Policy implementation is exceptionally difficult and not at all straightforward. Success in either depends critically on access to reliable and well-defined data. But high quality and clear definition, while necessary, are by no means sufficient to render data useful. The data also have to be relevant to the issues policy seeks to address and they have to be accompanied by a credible interpretive framework. These requirements, obvious though they may seem, are very difficult to satisfy. My remarks about these needs and difficulties are based on my own career as a scientist, administrator, and policy advisor, during which I have struggled with policy in nearly every stage of its complex life cycle from conception to death.