It is easily possible that all respondents answer a question with a scale from 1 to 5 with only 5 or only 1. This offers us very little or no information at all. Ideally, the answer distribution is normal, i.e. the majority of the answers is in the middle while the rest are distributed to the left and the right symmetrically.
This problem can often be predicted, either because we are familiar with such an issue or because we discover it during testing. We can solve it by adapting the question. If response values are too high, we make the question stricter by asking "I am extremely satisfied with XXX" instead of merely asking "I am satisfied with XXX". If response values are too low, we make the question more lenient, e.g. "XXX meets the most basic standards".