Firms redesign the evaluation form more often than they redesign anything else about the cycle, usually after a year when the ratings came back flat. The form is rarely the cause, but it is the most visible part and the easiest to change, so it gets changed.
It is still worth getting right. A form decides what a reviewer is able to tell you, and most of them ask for several things a partner cannot answer and leave out the field that does the work.
What the form is for
One job: to get a specific, attributable account of how an associate worked over a defined period, in a shape that can be compared with other accounts of the same associate and with the same account of other associates.
Everything on the form either serves that or takes up space a reviewer has limited patience for. Partners fill these in between other work, and a form that takes forty minutes produces either a late submission or a thin one.
The fields that belong
The working relationship, stated before the ratings. Which matters, roughly how many hours, over what period, and in what role. This is the field most forms omit and the one that makes the rest interpretable. A rating from a partner who supervised three hundred hours and a rating from one who supervised twelve are not the same evidence, and without this field they arrive looking identical.
Competencies with behavioral anchors. Not "judgment, 1 to 5", but a short description of what a 2 looks like and what a 4 looks like for this competency at this seniority. Anchors are what stop every reviewer using the scale differently.
A required example per competency rated at either end. If a reviewer gives a 5 or a 2, the form should ask for the instance. This single rule does more for rating quality than any scale redesign, because a reviewer who cannot think of the instance usually revises the rating.
Free text, with enough room. The written comment is where the information is. Scores tell you the distribution; the comment tells you what to do on Monday. Give it space and ask a specific question rather than offering "additional comments".
A development section with an owner. Two things to work on, and who is going to raise them. A development note with no name attached is a note nobody raises.
The self-assessment, collected separately and held. Not a field on the reviewer's form. The associate's own account belongs on its own instrument, collected first and released to reviewers only after they have filed, so a reviewer's rating is not anchored to what the associate claimed.
What to leave off
Billable hours as a rated competency. Hours are recorded elsewhere and they measure how much work an associate was given at least as much as how well it was done. Put them in the context pack if leadership wants them, not in the rating.
Personality adjectives. Proactive, team player, ownership, gravitas. A reviewer rating gravitas is rating something they cannot define and the associate cannot act on, and the ratings cluster because nobody wants to be the one who said a colleague lacks gravitas.
A single overall score. It gets used as a ranking, it is the first thing anybody looks at, and it is an average of things that should not be averaged. If leadership needs one number, derive it after calibration rather than asking reviewers to produce it.
Forced distribution, unless calibration follows. Telling reviewers that only 20% may be rated top band moves the problem from the rating to the allocation and tends to produce horse-trading rather than accuracy.
Anything a reviewer cannot exemplify. The test for any field: could a partner give an instance from the past year? If not, it will collect middle ratings from everyone.
The rating scale
Five points with written anchors works better than ten without them. A ten-point scale looks more precise and produces less spread, because reviewers cannot distinguish a 6 from a 7 and default toward the middle of the range they feel safe in.
Avoid an even number of points if you want reviewers to commit, and include an explicit "not observed" option that is not a midpoint. A partner who did not see an associate handle a client should say so rather than scoring 3, and if "not observed" is missing, the 3 is what you get.
A workable structure
- Reviewer and reviewee, the matters worked on, approximate hours, period and role.
- Six to eight competencies, each with anchors written for this seniority, each with a "not observed" option.
- A required example wherever a rating sits at either extreme.
- Two open questions: what should this person keep doing, and what is the one thing to change in the next six months.
- Two development actions with an owner and a revisit date.
That is a form a partner can complete in fifteen minutes, and it produces more than a longer one.
Before you redesign it
If the ratings came back flat, the form is probably not the reason. Flat distributions usually come from reviewers who are not sure the ratings will be used carefully, from a rater pool built on recollection rather than the matter record, or from a cycle with no calibration, so a generous reviewer and a severe one are never shown against each other.
Those are all fixable, and none of them is fixed by a new form. The sequence worth following is: get the rater pool right, add anchors and the required example, then calibrate, and redesign the layout last if it still seems to be the obstacle.
SRA designs the instruments it administers, including downward evaluations and 360-degree feedback, and will say which of the above is actually your problem before anything is rebuilt. The wider picture is in the associate evaluation guide.