For each project, consideration will be given during the design phase to the performance and impact evaluations that will be undertaken. USAID expects to devote approximately three percent (3%) of total program dollars, on average, to external performance and impact evaluation. This is distinct from resources dedicated to monitoring. While there is no hard and fast rule, USAID anticipates that roughly of 90% of the evaluations covered by these funds will be performance evaluations and the rest will be impact evaluations.
The Evaluation Component of a project MEL Plan covers impact and performance evaluations envisioned for a project. In addition, it may be useful to include any plans the Mission has for providing funds for self-evaluation under any of the implementing mechanisms through which elements of the project will be carried out. In a project MEL Plan, Missions may wish to include a Performance Evaluation Plan Summary table and a Project MEL and Learning Budget Worksheet as well as descriptions of the types of evaluations it contemplates. Templates for both of these tables are included in this section of the kit. Narrative descriptions in a project MEL Plan should provide a good overview of what is envisioned to serve as a starting point for the preparation of Evaluation Statements of Work, particularly for any impact evaluations that require an early start. Across this spectrum Evaluation Quality is Critical as illustrated by USAID criteria for quality evaluations in its Evaluation Policy and ADS 201.
Conventional wisdom holds that the quality of USAID evaluations steadily declined over the last two decades, with the majority of evaluations (with a few high-quality exceptions) relying too heavily on anecdotal information and expert opinion rather than on evidence collected in an unbiased and systematic way. While reviews of evaluations carried out by USAID found some exceptions to this characterization, most evaluators interviewed viewed evaluation quality as a critical issue.
—USAID Evaluation Policy: Year One (paraphrased)