Glossary
Interview rubric
An interview rubric describes what answers at each score level look like for a given question or skill, so different interviewers rate the same answer the same way.
Ask five interviewers what a “3 out of 5” means and you’ll get five answers. A rubric fixes that by writing down, in plain terms, what each level sounds like. It’s the difference between a rating scale and a shared standard.
An example
Skill: Clear explanation. Question: “Explain how to set up two-factor authentication to someone who isn’t comfortable with technology.”
| Score | What the answer sounds like |
|---|---|
| 1 | Uses jargon, skips steps, or can’t structure the explanation |
| 2 | Gets the steps right but rushes, with some unexplained terms |
| 3 | Short, ordered steps in plain language |
| 4 | Plain steps, checks understanding along the way, anticipates where the person might get stuck |
Each level describes something you can hear. None of them rely on words like “good” or “excellent.”
How to write one
- Collect real answers. Past interview notes or recordings are ideal. Otherwise, ask two colleagues to answer the question.
- Describe the best and worst first. The middle levels are easier once the ends are clear.
- Use observable behavior. “Gives a specific example” can be checked. “Seems confident” can’t.
- Keep it short. One line per level. Interviewers won’t read a paragraph mid-conversation.
- Match it to the level you’re hiring for. A rubric for an entry-level role shouldn’t expect expert answers to score well.
Calibrating
Once the rubric exists, test it. Have two or three interviewers rate the same three recorded answers independently, then compare. Where they disagree by more than a point, the wording is probably ambiguous. Rewrite it and repeat. Half an hour of calibration saves weeks of inconsistent ratings.
Rubric versus scorecard
A scorecard lists the skills and weights. The rubric sits underneath it and explains how to rate each skill. You can have a scorecard without a rubric, but the ratings will drift.
In AI-scored interviews
AI scoring needs the same thing a human does: a clear idea of what good looks like at the level you’re hiring for. When you set an expected level for a skill (beginner, intermediate, advanced or expert), you’re giving the model a rubric anchor. The more precisely the skill is described, the more consistent the scores.