What the Interviewer Is Scoring

Behavioural interviewing looks unstructured and usually isn't. Most companies of any maturity hand the interviewer a short list of attributes to rate with evidence, and the write-up afterwards is organised by those attributes rather than by your narrative. At the target companies the list is some version of:

Scope and ownership — how big was the thing that was genuinely yours, and did you stay with it past the merge. Judgement — did you make a real trade-off and can you name what it cost. Collaboration and conflict — what happens when someone disagrees with you, especially someone more senior. Drive through ambiguity — do you move when nobody hands you a spec. Communication — could a non-expert follow your explanation of your own work. Frequently a sixth: learning, meaning whether the story ends with something you changed about yourself.

Two consequences follow.

The first is that the interviewer needs evidence, not conclusions. "I'm very collaborative" rates nothing. "The staff engineer wanted to keep the sync API; I built both versions behind a flag and we compared the p95" rates collaboration and judgement in one sentence. Write-ups quote candidates; give them something quotable.

The second is that one story can cover several signals, which is why the story bank is small. A well-chosen story about a migration can carry scope, judgement, conflict, and communication at once. That's also why interviewers dig: the follow-ups are them hunting for the specific signal the first pass didn't cover.

Silence after your answer is usually note-taking, not disapproval. Let it happen instead of adding a fourth example.

Knowing the rubric changes what you keep and what you cut under time pressure. Most candidates over-invest in technical detail — which is scored in other rounds — and under-invest in the decision, the disagreement, and the outcome, which are the only things this round can measure. It also explains why great engineers get rejected: not because the work wasn't good, but because the account of it contained no evidence for the rubric.

• Assume a written rubric: scope, judgement, collaboration, ambiguity, communication, and often learning.
• Interviewers record evidence and quotes, so state the concrete decision rather than the self-assessment.
• Technical depth is scored elsewhere; here it's context, and too much of it crowds out the signal.
• Design stories to hit multiple signals; that's why eight good stories cover a whole loop.
• Follow-up questions are signal-hunting, not scepticism — answer the specific thing asked rather than restarting the story.
• End on what changed as a result, for the system and for you; the learning signal is cheap to earn and often skipped.
• Ambiguity has its own line item: say what you did when nobody told you what to do.