Score candidates on evidence, not vibes
Unstructured interviews are notoriously poor predictors and notoriously easy to bias. A rubric fixes both problems by naming the skills you care about and defining, in advance, what each score looks like in observable behavior. This builder turns your competencies and rating definitions into a clean, shareable rubric your whole panel can apply the same way.
How it works
You list the skills (competencies) you assess and, for each, write behavioral indicators for a 1 to 4 rating. The tool assembles them into a Markdown rubric with a row per skill and a column per level, plus a calibration section. The even 1-4 scale is deliberate: it removes the comfortable middle and maps cleanly to:
- 1 — strong no: clear gaps; would struggle in the role.
- 2 — lean no: below bar; notable concerns.
- 3 — lean yes: meets the bar; solid evidence.
- 4 — strong yes: clearly exceeds; standout signal.
You can mark some skills as higher weight when they matter more for the role. Everything is generated locally in your browser.
Tips and example
Write indicators as things you can see, not adjectives. Instead of “good problem solving,” a level-3 indicator might read “breaks the problem into sub-cases, states assumptions, and tests against an edge case.” A level-1 might read “jumps to code without clarifying the requirements.” Concrete language is what makes two interviewers agree.
Keep the rubric to a handful of skills that genuinely predict success in the role; a rubric with fifteen competencies dilutes attention. Weight the few that matter most so a strong system-design signal is not cancelled by a weak opinion on naming style.
Calibrate before you interview. Have the panel score one or two recorded or sample answers together and reconcile differences so everyone shares the same bar. Revisit calibration periodically, because rating standards quietly drift as a team interviews more candidates.