I think it is dangerous to be too qualitative or too simplistically quantitative about judging a candidate's accomplishments. I do believe, however, that looking at Mathscinet citations is a good start in evaluating both a person and the person's papers. Basically, you want to find evidence that a person's papers have had good impact in the sense that it has led to other respected work (as just by, say, the quality of journals and number of citations of *those* papers) that use or build on the person's contributions.

All of this is quite problematic when you are judging someone who has been publishing for less than 10 years, but in my experience, for someone out 10 years or more, looking at citations per paper or citations per year, as well as total number of citations, is a remarkably good guide for identifying the better mathematicians, in the sense that the ranking I get agrees rather well with my own subjective views. Then I simply focus on the exceptions and try to decide whether the citations are telling me that I've misjudged or whether the citations are simply misleading in that particular case. Even there, my conclusion is usually but not always the former.

But there are still difficulties. It is quite noticeable, even within pure mathematics, that people working in some fields (like PDE's) get a lot more citations than others. So, you have to be careful about comparing people across different fields.