Although it is sometimes necessary to develop a quick survey, such as the first one TIS conducted in 1990 (see Appendix I), the best results come from a carefully designed and professionally produced
survey. Working with the survey designer, list all the data and elements that would be important to learn from the survey results. Group the questions by subject and type. Then eliminate all but the essential questions. Draft the survey beginning with easy factual information, and progressing to questions flowing logically from one point to the next. Define auy terms the clients may not know. For example, librarians think of document delivery while clients may think of article orders. Use a variety of different question types, but not so many that a respondent might become co~~fused or the results might be so varied as to be difficult to interpret or combine. Avoid too many open-ended questions since they are more difficult to tabulate, but provide space for comments. Remember that the survey is one for busy business people, not for members of an academic community more attuned to the role of questionnaires in evaluating services.