The first ten interviews I ran, I asked whatever I felt like that morning. Same role, different questions every time, and no record of what a "good" answer had looked like two candidates ago. When the team compared notes afterward, we realized we weren't comparing people β we were comparing the days we'd had.
So I started generating question sets with a scoring rubric attached to each one. The rubric is the product, not the question.
Why a rubric per question changes the conversation
A question without a rubric is just a conversation. "Tell me about a time you shipped something late" β fine question, but three interviewers will read three different answers as "strong." The rubric forces you to decide before the interview what a strong answer contains: did they name the trade-off, did they own the miss, did they say what they'd change. Now the score is about the answer, not the interviewer's mood.
The seniority tuning matters here too.β¦