This is written from an evaluator's perspective, but I address those who commission evaluations, too:
- Cluster and prioritise the questions and draw a sample of projects, in dialogue with the client. This may sound banal - but anyone who has tried knows how hard it can be to get people to shed some questions. Even with a clear sense of priority, you won't get "scientific evidence" within that time frame, but the fewer questions you work on, the deeper you can go. Gathering data against, say, 30 evaluation questions on 40 projects run by 30 organisations would mean asking thousands of questions, also because it is not enough to gather data from just one source per project. But you can fit only about ten questions into an hour's semi-structured interview or a couple of dozens into a ten-to-fifteen-minute standardised questionnaire.
- Work in tandems! Well-informed external opinions are even more valuable if they come from more than one person. In two of "my" recent quick evaluations, I - the evaluation specialist - was paired up with a regional specialist. We did some work together, some in parallel, and discussed our findings every day - the result has been so much richer than what a single consultant can produce in so little time. If the budget does not allow for joint field work, you can still organise a desk-based peer review which would mean just an extra couple of days' honoraria.
- Put in three more days on site. Whatever the original "quick" time frame is, three more days make a huge difference. You need a day for inception on-site and at least a day, better two, to analyse initial findings and discuss them with the people who run the programme. That will take two to three days off your week - adding just three days to the overall exercise will double the time for data gathering and analysis.
- Mix your methods, even if there is no time for "proper" sampling and statistics. If your main instruments are interviews and group discussions, look at the budgets and the spending as well - seeing where the money goes tends to yield valuable insights, too.