This paper discusses some of the major issues relating to the planning of evaluations of federal human services national demonstration programs. It is argued that there is often insufficient planning and emphasis on evaluation, which limits the amount of useful knowledge obtained by these demonstrations. Barriers to conducting evaluations are identified for the policy level and cross-site and within-site evaluation contexts. Some of the proposed solutions include utilizing an explicit programmatic developmental sequence to evaluation pegged to program stages and modifying the role and functions of cross-site evaluators.