I also think stack-ranking is hard or even impossible to get right. The criteria often favor those people that produce visible work and tend to disfavor people that enable said work (think: producing new features vs. maintenance, writing high-profile code vs. doing the review for those, producing visible output vs. mentoring, ...)

So stack ranking (or any kind of ranking by criteria) tends to nudge people into "let's do what the standard expects us to do" instead of "let's do what makes the team, company and product improve."