Because the evidence is gathered in the final week, from memory, which is heavily weighted towards the last two months. Recency bias is not a flaw in the reviewer, it is a consequence of the process design.
A year assessed from three weeks of recollection is a review of the third quarter with an annual label.
Short, dated notes through the year — a paragraph after a significant piece of work — cost almost nothing and remove the recency problem entirely.
This is the single highest-return change to a review process, and it does not require a new framework or a longer form.
Continuous evidence with a periodic conversation. The annual assembly of memory is the part that fails.
Only with real calibration across reviewers. Otherwise they are noise with a number attached.
Separately. Together, the pay question crowds out everything else.
Half an hour with your own data usually saves reading three of these. The guides will still be here afterwards.