Assessing Someone You Will Never Meet

The interviewer's confidence in a video interview is roughly the same as in person. The accuracy is not. Knowing which signals you have lost is most of the fix.

What the screen removes

Physical presence and its noise. You no longer register height, handshake, posture or dress in the same way. Much of that was bias rather than signal, so losing it is partly a gain — but it also means your intuitive sense of "fit" is now running on less input while feeling equally certain.

Peripheral behaviour. How someone interacts with people who are not interviewing them.

Environmental context. You see a background they curated. A candidate in a shared flat with poor lighting and one in a home office tell you about housing, not competence — and it is easy to read one as more professional than the other.

That last point is worth stating plainly: assessing a candidate's home is assessing their income.

What to substitute

Written work. The most reliable substitute and the most neglected. Give a short prompt requiring a structured answer. In distributed work, clear writing is not a nice-to-have; it is the medium.

A working session rather than an interview. Ninety minutes on a real problem, together, with them driving. You learn how they think, what they ask, how they handle being wrong — most of what the room used to tell you.

Deliberate behavioural questions. Not "tell me about a challenge" but questions with a specific shape: describe a time you were blocked and the person who could unblock you was unavailable. What did you do in the first hour, and in the first day?

A conversation with a non-assessor. Someone on the team, no scorecard, both directions. It gives the candidate a real view and gives you a signal you will not get from a structured interview.

Structure protects you here

Unstructured interviews are weak predictors in person and worse on video, because the intuition they rely on has less to work with.

Same questions, same order, for every candidate. Boring and it works.

Score immediately after each conversation, before discussing with anyone. Group discussion first produces convergence on the loudest opinion.

Assign each interviewer one dimension. Diffuse responsibility produces four people all forming a general impression.

Write down what would change your mind before the interview. It makes confirmation bias visible.

The specific traps

Judging the setup. Lighting, background, headset quality. This is a proxy for money.

Penalising connection problems. Broadband quality is geography and income, not diligence.

Reading camera-off as evasion. People have reasons — bandwidth, environment, disability, a sleeping child.

Over-weighting fluency. Video amplifies the advantage of people who present smoothly. Ask yourself whether you are assessing the answer or the delivery.

Assuming quiet means disengaged. Interrupting is harder on video, so people wait. Some of the best candidates seem passive because they are being polite about a broken medium.

Reference checks matter more

With fewer direct signals, secondhand ones carry more weight.

Ask specific questions rather than general ones. "Would you work with them again" is nearly useless — almost everyone says yes. "What did they need most support with?" and "how did they handle it when they disagreed with you?" produce information.

Tell them what you decided and why

Remote processes feel impersonal to candidates because they are. A rejection with two sentences of substance costs you five minutes and is remembered for years — including by the people they talk to.