It depends. Ranking produces sharper distinctions, but it is more difficult for respondents to complete. Respondents may also produce artificial distinctions. Ratings are easier to do but they may not make a sufficient distinction between items and are influenced by response style and by social desirability.
Comparisons of ratings and rankings show that average preference orders are similar, but individual preference orders are less similar. A useful solution is to use both scales in combination: ask for ratings and then ask for the three most important.

