Mutually Assured Mediocrity

One of the best tools for peer review is forced rankings. Asking questions about a peer's performance will typically get you either a bland non-answer or a heavily biased opinion that has little to do with the specific performance you're actually asking about - it's just the nature of interpersonal relationships, we don't view Bob's communication skills as a separate thing to be analyzed independently of our opinion of Bob as a person. Forced rankings where a person's performance at one specific task must be compared to multiple other peoples', and ties are not permitted, gives us the freedom to disassociate the skill from the person and treat it more objectively. It's also more easy to compare responses from multiple sources. Bob may be a nice guy who is always happy to lend a hand, but if he is still the last person anyone on the team would go to for an explanation of a tough technical concept, you know what he needs to work on.