From a statistical standpoint, try to keep as many of your questions in close-ended categories that you can assign point values to.
For example, I use often use these standard response categories in agreeing/disagreeing with a given statement: strongly agree (4), agree (3), neither agree nor disagree (2), disagree (1), strongly disagree (0). You get the idea. Just keep all your "high scores" in the same direction, so that, when you average the survey results, higher point values indicate higher customer satisfaction and vice-versa.
When I do the close-ended questions, this makes collection, compilation, and reporting of the data *much* easier.
Sure, I throw in 1-2 open-ended questions at the end to let people vent and collect new ideas, but if you're going to use this for benchmarking, you need to have enough close-ended questions with point values to give you hard numbers. It is often nearly impossible to point-scale open-ended responses, because the evaluator's take on them is always subjective. I usually just pick through these and use them as illustrative points in the final report.
If you stick to discrete response categories, then you can easily show measurable improvement (or decline, whichever the case may be) in a much easier fashion over time. It's soooo much less painful to do it this way, especially if you're dealing with a large group of victims/respondents.
