How to Score Candidates Fairly After Interviews

You've just finished four interviews for one role. Candidate two was warm and quick on her feet. Candidate three gave flatter answers but every one of them was backed by something she'd actually built. You liked candidate two more. You're not sure that means anything.

This is the moment scoring is supposed to fix, and it's also the moment most scoring falls apart — because the scorecard gets filled in after the chat, from memory, shaped by whoever talked last or made you laugh. A scorecard only works if you build it before you meet anyone and use it while you're still in the room, not as a summary afterwards.

Why "I'll just compare notes" doesn't work

When a hiring team meets to "discuss impressions," the loudest opinion in the room usually wins, and it tends to be whoever interviewed the candidate who reminded them of themselves. Nobody is lying. Everyone genuinely believes their read. That's the problem — unscored comparison feels objective from the inside and isn't.

The fix isn't to trust your gut less. It's to write down what your gut is reacting to, against the same yardstick, before you compare notes with anyone else.

Build the rubric before the first interview

Pick 4–6 things this specific role actually requires. Not "communication skills" in general — the version of communication this job needs. For a customer support lead that might be "explains a technical problem to a non-technical person clearly." For a sales role it might be "handles being told no without getting defensive."

For each one, write three short anchors — what a 1, 3, and 5 actually look like. This is the part everyone skips and it's the part that matters most, because "rate communication 1–5" means something different to every interviewer. "Explained the refund policy so a non-technical person could repeat it back" is a 5. "Used jargon without checking understanding" is a 2. Now two interviewers scoring the same answer land in the same place.

Example rubric line:

Criterion: Handles a customer complaint calmly
- 1 — Got defensive or blamed the customer in the example given
- 3 — Stayed polite but didn't actually resolve anything
- 5 — Walked through a specific complaint, named what the customer needed, and how they got there

Do this for every criterion, print it, and hand a copy to every interviewer before they walk in.

Score alone, before you talk to anyone

This is the single change that fixes the most bias for the least effort: every interviewer fills in their scorecard and writes one line of evidence per score, alone, within an hour of the interview — before the group debrief, before Slack messages, before "so what did you think?"

The evidence line matters more than the number. "4 — gave a specific example of talking down an angry customer, named the exact words she used" is defensible. "4 — good vibe" is not, and forcing yourself to write the evidence often changes the number you'd have given from memory.

Weight the criteria, don't just average them

Not everything on your list matters equally. If you're hiring a bookkeeper, accuracy under pressure probably outweighs how personable they are in the interview. Decide the weights before you see anyone — 40% technical accuracy, 30% attention to detail, 20% communication, 10% culture fit, or whatever your role needs — and do the multiplication, don't eyeball it. A candidate who's a 5 on the thing you weighted heaviest and a 3 on the thing you weighted lightly should usually beat a flat 4 across the board, and a weighted total makes that visible instead of argued about.

What to do when two candidates are close

Sometimes the rubric genuinely produces a near-tie, and that's fine — it means both candidates are plausible and the interview format has told you what it can. Don't resolve a close score with another round of "who did we like more." Resolve it with something that produces new evidence: a short work sample task both candidates do under the same conditions, or a reference call that asks about the specific thing separating them. If the gap is in a skill you can measure directly — a numeracy check, a written response to a realistic scenario — a structured test with a published difficulty baseline, like the item bank AssessFit runs its tests against, gives you a number instead of a second impression to argue about.

Where scoring still doesn't remove bias

Be honest with yourself here: a scorecard formalises your judgement, it doesn't purify it. If your anchors reward polish and confident delivery, you'll still favour candidates who interview well over candidates who do the job well — you'll just have a spreadsheet saying so. Write anchors around specific, observable behaviour rather than impressions ("gave a concrete example" rather than "seemed confident"), and get a second interviewer to score the same interview independently once in a while, just to check your anchors mean the same thing to both of you. If two trained interviewers watching the same answer land three points apart, the rubric needs rewriting, not the candidate.

A scorecard you can use tomorrow

1. List 4–6 role-specific criteria.
2. Write a 1/3/5 anchor for each, in plain observable language.
3. Assign a weight to each criterion that adds up to 100%.
4. Every interviewer scores alone, with one evidence line per score, within an hour of the interview.
5. Multiply scores by weights before anyone discusses impressions.
6. If it's close, get new evidence — a task, a reference call, a second scored interview — don't argue it out.

None of this makes the decision for you. It just means when you pick candidate three over candidate two, you can say why in a sentence that isn't "I had a feeling," and that sentence will still make sense to you in six months when you're wondering whether the hire worked out.