Time and money limitations normally make it impossible to do usability testing on every document, particularly by the method of controlled observation of users performing set tasks. I describe approaches that are intended to make the feedback from testing useful to more writers than just those who wrote the tested documents, and that gradually improve the corporate wisdom available to writers. Methods of approaching these goals include judicious selection of what to test and how, special empirical studies on issues of general importance, improving the quality of a company's prescriptive rules, and better ways of making new knowledge known and available to writers.