A structured interview is only as good as its rubric. If two interviewers can score the same answer differently, you do not have a structured interview, you have a conversation with numbers on it. This guide shows how to build a STAR-method scoring rubric that is fair, consistent and defensible.
STAR in one minute
STAR stands for Situation, Task, Action, Result. A complete behavioural answer names a specific situation, what needed doing, what the candidate personally did, and what happened as a result.
The rubric template
Pick 4 to 6 skills that matter for the role. For each, define the skill in one sentence, a weight, and what a 1, 3 and 5 look like.
Template
Skill: Incident response (weight 25) Definition: Contains and resolves a security incident, and communicates clearly under pressure.
- 1: Generic answer with no specific example; no clear personal actions.
- 3: A real example with a sensible action; limited detail on decisions or outcome.
- 5: A specific, recent example; clear personal ownership; sound decisions and trade-offs; a measurable result; learning applied afterwards.
Repeat for each skill, then add the weights to 100.
The scoring rules that make it fair
- Score the evidence, not the style. An accent, a nervous start or a typo is not evidence.
- One question, one skill. Mixed questions muddy scoring.
- Evidence first. Write the supporting quote before the number.
- Ask the follow-up. If the answer is vague, ask what exactly the candidate did and what happened as a result.
- No halo effect. A strong first answer does not lift the next skill.
- Calibrate. Two reviewers score the same three answers and compare.
Write STAR-friendly questions
Weak: "Tell me about your experience with detection engineering." Better: "Tell me about a detection you built or tuned that cut noise or caught something real. What was the logic, and how did you measure it?"
The second prompts a situation, an action and a result. Add a follow-up hint for the interviewer: "Ask how they measured the improvement."
How an AI interview uses the rubric
In Critical Hive you paste a job description and the platform drafts competencies and STAR-style questions that you edit freely. Each answer is scored 0 to 5 per skill, with the candidate's own sentences as evidence and a plain-English reason. A person can read the evidence and override anything.
Calibrating AI scores
- Run ten past candidates you already know through the rubric and compare.
- Look for skills where the AI is consistently higher or lower than your people.
- Fix the rubric text before you blame the tool.
- Re-run after any change.
Common mistakes
- Too many skills: more than six dilutes scoring.
- Weights that ignore the job: a customer-facing role should weight communication.
- No definition of "5": if a 5 is vague, everyone gets a 4.
- Treating the score as the decision.
FAQ
How many questions should an interview have?
For a first round, around five well-written questions with short follow-ups gives enough evidence without exhausting the candidate.
Should I weight skills differently?
Yes, if the job does. Document why, and keep the weights visible to reviewers.
Is a 0 to 5 scale better than 1 to 10?
Fewer points are easier to define and score consistently. A five-point scale with written anchors is a good default.
Next step
Download the presentation for a walkthrough of the scoring and evidence screens, or try it free on one role.