Endoscopic therapy clearly has a primary role in many clinical conditions (such as dysphagia, obstructive jaundice, bleeding, and colonic polyps). There is much less certainty about the role of endoscopy in many other clinical contexts (e.g. management of chronic pancreatitis). Randomization is the gold standard for evaluation of competing therapies. Unfortunately, there are many difficulties in mounting meaningful randomized controlled trials of endoscopic methods. Many have been done, but few have provided us with real practical answers. This article argues that the evidence we need to advise patients is often better obtained through very stringent observational studies, provided that all necessary data elements are defined and documented and independent objective arbiters (referees) are fully involved in the process.