How to build an interview scoring rubric with behavioural anchors
Rating “communication” out of five tells you nothing, because nothing defined what a four was. How to write anchors an interviewer can actually match an answer against.
A behaviourally anchored rating scale replaces abstract qualities with observable descriptions at each level, so an interviewer matches what a candidate said against a description instead of converting an impression into a number. Write the extremes first, describe what the candidate says or does rather than what they seem like, mark the genuinely must-pass dimensions, and keep the whole rubric to four to seven dimensions.
Why most scorecards do nothing
The standard interview scorecard asks for a rating from one to five on a handful of qualities, with no definition of what any number means. Two interviewers watch the same candidate and one gives a 3 while the other gives a 5, and neither is wrong, because nothing said what a 4 was.
The scores then get averaged, which lends the whole thing an arithmetic respectability it has not earned, and the decision gets made in the debrief on whoever argues most confidently.
What a behavioural anchor is
A behaviourally anchored rating scale replaces the abstract quality with observable evidence at each level. Instead of rating “communication” out of five, you describe what a weak answer sounds like, what an adequate one sounds like, and what an exceptional one sounds like, for this role.
The interviewer is then matching what they heard against a description rather than converting an impression into a number. That is a far easier task and a far more repeatable one.
| Level | Unanchored scorecard | Behaviourally anchored |
|---|---|---|
| 1 | Poor | Describes the outcome only. Cannot say what they personally did or what the alternatives were. |
| 3 | Meets expectations | Describes their own contribution and one trade-off they weighed, with a concrete example. |
| 5 | Outstanding | Names the decision they got wrong, what it cost, and what they changed as a result. |
An interviewer matching an answer against a description is doing a different job from one converting an impression into a number.
Writing them
- Start from the work, not the quality.Not “problem solving” but “diagnosing a production issue with incomplete information”. Anchors need something concrete to describe.
- Write level 5 first, then level 1. The extremes are easier and they set the range. The middle is then interpolation rather than invention.
- Describe what the candidate says or does.If an anchor contains a judgement word like “strong” or “good”, it is not an anchor yet.
- Mark the ones that are genuinely must-pass. A failed must-pass should pull the overall result down regardless of the average, or the average will quietly rescue a candidate who failed the thing that mattered.
- Keep it to four to seven dimensions. Beyond that, interviewers stop reading the anchors and revert to impressions, which puts you back where you started.
The failure modes
- Anchors that describe the person rather than the answer.“Confident communicator” is a personality judgement and is exactly where bias enters. “Explained a technical trade-off to a non-specialist without jargon” is observable.
- Reusing one rubric across levels. The anchors for a first-time manager and a director are different documents. Sharing them means one of the two is being measured against the wrong bar.
- Scoring what was never asked. If the interview did not probe something, the honest record is that it was not probed. A zero is a claim about the candidate; silence is a claim about the interview.
- Writing them after the interviews start. A rubric written once you have met candidates is a description of the people you liked.
The test
Hand your rubric to someone who has never met the candidate, with a transcript. If they arrive at roughly your scores, the anchors are doing their job. If they cannot, the document is decoration and the decision is still being made on impressions.
Common questions
- What is a behaviourally anchored rating scale?
- A scoring scale where each level is defined by observable behaviour rather than an abstract label. Instead of rating communication one to five, each level describes what a weak, adequate and exceptional answer actually contains for that role.
- How many dimensions should an interview rubric have?
- Four to seven. Beyond that interviewers stop reading the anchors and fall back on impressions, which removes the benefit the rubric existed to provide.
- What makes an interview anchor biased?
- Describing the person rather than the answer. Phrases like confident or articulate are personality judgements and are where bias enters; an anchor should describe something the candidate said or did.
- How do you know if a rubric is working?
- Give it to someone who never met the candidate, along with a transcript. If they reach roughly the same scores, the anchors are carrying the decision. If they cannot, the rubric is decoration.
Related reading
All posts →See it on one of your own roles
Upload a job description and watch the pipeline run before you commit to anything.
Start for free