A useful no-nonsense (if ever so slightly dated) guide on all the things to consider when running a randomized field experiment for a social or economic program, from some of the pros at J-PAL. The target is aid practitioners, and while it covers many of the topics an academic analyst would consider (like analysis of power, spillovers, noncompliance, loss to follow-up, etc), quite a bit of it is focused on less academic but crucial issues on the project management side, and is insightful about the interaction of the many many practical implementation constraints (getting buy in from local partners and governments, designing intervention and data collection protocols implementable by local staff, ethics, ethics, ethics, and more ethics, etc) with the design of an experiment which will actually be informative and drive policy decisions. It really makes clear that 90% of the work of running experiments is outside the basic "flip a coin and then see if the thing works" that you see in the typical academic writeup of one of these studies.
Since its 2013 publication, best practices for data analysis have evolved somewhat (so consult your local expert for advice on incorporating baseline covariates or clustering standard errors), and more experiments are designed with testing social scientific theories in mind, which engages a somewhat different, though complementary, set of concerns, but I would still classify this as required reading for anyone about to embark on running such an experiment, and important background for anyone looking to understand and engage critically with experimental literature in this area.