Interview scorecard template: what a four out of five hides

An interview scorecard template is only useful if every rating points back to something the candidate said about real work. A grid with "communication," "ownership" and a score from one to five can still hide a gut feeling if nobody has defined what the numbers mean. (Ask three interviewers what a four means and you will get three answers.)

So build it backward from an actual task: the capability it requires, the question that could reveal it and the behavior that justifies each rating. The example below is a starting point to review with people who know the job, not a validated selection instrument.

1. Start with a task, then write the question

OPM's job-analysis guidance links work tasks to competencies, and its structured-interview guide turns that link into questions and rating scales. Read the job-analysis overview and the interview guide.

Here is a made-up example for a customer implementation role:

Link	Worked example
Task	Pass a customer issue to the right internal owner with enough information to act
Capability	Clarifies missing information and makes ownership explicit
Question	Tell me about a handoff you managed when the next step or owner was unclear. What did you do, and what happened?
Neutral probes	What did you know then? What was your own contribution? How did you determine whether the handoff had worked?
Evidence recorded	Candidate actions, constraints, follow-through, result, and unresolved questions

The question never asks whether the candidate is "good at ownership." It asks for an example you can examine.

2. Write the rating anchors before anyone interviews

For this one task, a three-level scale could look like this. It is deliberately coarse: adding numbers without real distinctions makes a scorecard look more precise than it is.

Rating	Evidence demonstrated in the answer
1: below this role's requirement	Describes forwarding the issue but, after standard probes, gives no clear method for resolving missing ownership or needed information
2: meets this task requirement	Explains how they identified the next owner, assembled the needed context, and checked whether the handoff progressed
3: strong evidence for this task	Meets level 2 and explains a relevant failure or constraint they detected, how they adapted, and what evidence showed whether the adaptation worked
N/A: not assessed	The question was not covered, the record is inadequate, or the permitted assessment did not provide enough evidence to rate it

A rating describes the evidence from this assessment, not the person's permanent ability. A failed project can still show strong judgment.

The OPM guide recommends behavior examples to separate proficiency levels, but as guides, not scripts every candidate must reproduce. Someone may show the same capability through an unfamiliar but relevant approach.

3. Write the evidence next to the rating

A made-up assessor note might read:

Paraphrase: The candidate described a customer issue that initially had no internal owner. They separated the missing technical details from the approval question, confirmed a responsible person for each, and checked the next day that the technical investigation had begun. The customer deadline was still missed because approval took longer than expected. The candidate did not describe a change to the future handoff process.

Under the example anchors, that is level 2. The missed deadline alone does not make it level 1, and the answer lacks the adaptation level 3 asks for.

When assessors disagree, name the anchor or fact in dispute. "I liked the answer more" is not an evidence standard.

4. Treat a skipped question as missing, never as zero

Do not enter zero for a skipped question, average N/A into a total or let a high score elsewhere cover a missing required capability. Decide in advance what happens instead: an equivalent follow-up assessment, a process correction, or a decision that stays open.

If a task has to meet a threshold, state that rule before interviewing so a total cannot hide it. Any weighting needs a job-related reason; decimal places do not validate it.

5. Test the scorecard on made-up answers first

Write two or three made-up answers with different strengths, have interviewers rate them separately, and compare. If they split because "checked progress" means different things to each person, fix the anchor before using it with candidates.

Also check that the question fits the time and the probes clarify without handing over the answer. The OPM guide includes piloting as part of interview development.

Version the scorecard and note why each real change was made. If a change alters what counts as enough evidence, consider its effect on candidates already assessed.

Check the anchors with someone who does the job

The anchors are only as good as your picture of the work. If nobody on the team has done this job, a short call with someone who does it at another company can tell you which handoffs actually go wrong and what a strong answer would contain. On Instant Expert you describe the role, such as "implementation managers at B2B software companies," set what each person receives for a 30-minute call, and pay our fee on top only if they book: a $100 offer costs $125. Ask them to rate your made-up answers from step 5 against the anchors, the same test your interviewers take. Say in your note that the call is about the scorecard, not a job opening, so nobody mistakes it for recruiting.

Start with the one task this hire most needs to do well, and write its anchors before the first interview is booked. A score written after the team has picked a favorite is decoration.

Interview scorecard template: what a four out of five hides, and how to build one from real work