Accept in good faith that the author(s) have laboured long and hard over the paper (unless there is evidence to the contrary), and that someone's career partially hinges on your review to their paper. Then do the following:
1. Scan the paper once quickly, making notes on the central thesis and evidence for it.
2. Have a long hard think about the paper for at least five days without reading it again (resist the temptation to fire back an indelicate or hasty review - This is particularly important if you know the area well).
3. When you have collected your thoughts, read the paper again slowly and dispassionately, making further notes.
4. Write your review beginning with positive points about the paper. Then move to the more doubtful claims or evidence noting clear points of dispute or rebuttal. Other responders have noted the key things to look out for so I won't repeat them.
5. Wrap up the review making constructive suggestions as to where the author can lift their game. Include references to key works that might have escaped the writer's notice.
We have all, at one point, been at the mercy of petulant reviewers and it behooves us to avoid doing the same when it is our turn.
Sorry, "review" could be ambiguous in this case --- I assume you are talking about refereeing a paper for publication. In that case, generally, if you are reviewing for a journal or conference there are guidelines that are given to the authors and reviewers and you don't really have a choice. But, even within those guidelines what is important to emphasize requires some judgement in terms of how you can best help the authors.
If you instead mean writing an article that is a review of another, then I would look for guidelines for review articles in journals of your subfield. What is important depends on your perspective. Often a review article of an individual paper will place the results in a broader context, while an article that surveys an area will use different papers to tell a broader story than is given in those individual papers. In this case, the answer is to emphasize what helps support the points you are trying to make.
Accept in good faith that the author(s) have laboured long and hard over the paper (unless there is evidence to the contrary), and that someone's career partially hinges on your review to their paper. Then do the following:
1. Scan the paper once quickly, making notes on the central thesis and evidence for it.
2. Have a long hard think about the paper for at least five days without reading it again (resist the temptation to fire back an indelicate or hasty review - This is particularly important if you know the area well).
3. When you have collected your thoughts, read the paper again slowly and dispassionately, making further notes.
4. Write your review beginning with positive points about the paper. Then move to the more doubtful claims or evidence noting clear points of dispute or rebuttal. Other responders have noted the key things to look out for so I won't repeat them.
5. Wrap up the review making constructive suggestions as to where the author can lift their game. Include references to key works that might have escaped the writer's notice.
We have all, at one point, been at the mercy of petulant reviewers and it behooves us to avoid doing the same when it is our turn.
All the answers above make for good guidelines for reviewing. Stefan and Martin present viable, specific areas of focus and their lists should definitely be taken into consideration.