The article by Ricki Lewis, "Teaching Evaluations: Widespread And Controversial" (The Scientist, 12[9]:12-13, April 27, 1998), hit some important points regarding the use of teaching evaluations. I suggest two other (uses of evaluations):
Daily evaluations: If the purpose of teaching evaluations is to improve teaching, I recommend instituting daily evaluations of every lecture, rather than taking them when the semester or quarter is over. For the evaluations to have the greatest impact, the improvements must be made during the time that it can help the students writing the evaluations. I do this and find the process simple and confidential. At the beginning of every class, I pass out 10 to 20 half-sheets of paper with a few questions, such as, "Did you find any parts of the lecture unclear?" or "Did I cover the material too slowly, too rapidly, just right?" Then I leave space for comments. Usually...