You may have heard recently about the idea to use randomized field trials to evaluate the success or otherwise of develoment interventions. It's a great idea to add rigour to the evaluation process and there's obvious potential here for finding out (to use a horrible phrase) "what works".
But the problem is this. If development agencies and donors are not forced to publish their results and be transparent about their failures then the lessons from set-backs will not be identified, let alone shared. At the moment they face no incentive to let the world know about their failures because, they imagine, this would only lead to reduced donations and funding.
As ever, political factors get in the way of progress.