Being a reviewer is no joke. The journal expects its reviewers to possess hawk-eyes in identifying minute errors and make sure the article is plagiarism free and also corresponds to the journal's entry requirements.
Assuming the reviewer is a subject matter expert, the most important first step is to understand several critical issues:
1) Does the article raise a grounded, reasonable, and important issue(s) in the field?
2) Are the methodology, tools, analyses, results discussion, etc. up to the field's best standards?
3) Whether results obtained are important indeed, or constitute at least a step forward sufficient to be considered for a publication?
The second step for the reviewer is to decide whether the found above unintended errors, mistakes, places having a lack of clarity, etc., if cleaned up could make the article publishable, or the level of deficiencies is beyond repair.
The third step is to check the article on compliance to the Journal/Editor requirements checklist.
The fourth step is to decide on whether to recommend, decline, or open a communication channel with the author(s), providing a formal review with a request for corrections.
Look for novelty, research methodology, compliance with the format and style of the journal, results and discussions, comparison with other researchers' works, etc.
In agreement with the above answers, reviewing should not be a traditional process for all kinds of papers. However, some basics must be conserved like research idea and goal, methodology and treatment of results. Discussion and conclusions must be considered also. Reviewers aim should not be rejecting articles but to correct it as possible.
i think that answer above is complex and dynamic, but the most of important in making decision of literature review or critical review, there are (1) problem statement, is suitable or not, related with the field problem, (2) Methodology, (3) result and discussion, and (4) conclusion and forth, then looking for weakness or strongest, and of course based on theory and field data...