Interviewing and Evaluating Candidates

Give every interview one assigned signal, require written evidence before anyone discusses the candidate, and ask about specific past decisions rather than hypotheticals. Most interview failure comes from several people forming a general impression in parallel and then converging on whoever was most personable.

One signal per interview, written before the debrief

The default loop has four people each forming an overall view. That produces four correlated impressions rather than four independent pieces of evidence, and the debrief becomes a negotiation between confidence levels.

Assign each interviewer one thing to establish and hold them to it. One person on technical depth, one on how the candidate handles ambiguity, one on collaboration, one on the specific domain the role requires. Each writes their evidence and their conclusion before hearing anyone else's.

Writing first matters as much as the assignment. Once someone has heard a colleague call a candidate impressive, their own recollection reorganises around it.

Ask about decisions, not approaches

Hypothetical questions produce rehearsed answers. Every experienced candidate can describe how they would approach an ambiguous project, and the answer tells you they have been interviewed before.

Past decisions produce evidence, because they have specifics that can be probed. What was the situation, what did you choose, what did you not know at the time, what happened, what would you do differently. The follow-up questions are where the signal is, which is why this format is difficult to prepare for.

The strongest single question remains asking what they got wrong. Candidates who cannot answer it either have not reflected or are managing the interview, and both are informative.

Testing for ambiguity directly

Give an underspecified problem and observe what happens in the first two minutes.

Commit-then-verify looks like: "I am going to assume this is for internal users, which means I would optimise for X. Tell me if that is wrong." Ask-until-clear looks like a series of clarifying questions with no interpretation offered. In a company with product managers and specs, the second behaviour is diligent. In a seed-stage company it is a dependency nobody can service.

Keeping the bar when you stop interviewing everyone

This is the transition that quietly lowers hiring quality, and it happens without a decision being made.

While the founder interviews everyone, the standard lives in their head and gets applied consistently. The moment they stop, the people hiring in their place are guessing at what the founder would have decided. Drift is not a failure of care, it is a failure of documentation.

The replacement is a written standard, with worked examples of a strong hire and a rejected candidate, plus one trained interviewer per loop who owns the bar. That role is not the hiring manager. It is someone whose job in the loop is consistency across candidates rather than enthusiasm for this one.

Common questions

How do I interview candidates when I've never done it before?

Give every interview one assigned signal and require written evidence before anyone discusses the candidate. Most first-time interviewing failures come from several people forming a general impression in parallel, then converging on whoever was most personable rather than most qualified.

What questions should I ask candidates that actually predict success?

Questions about specific past decisions rather than hypotheticals. "Tell me about a time you had to ship without complete information, and what you got wrong" produces evidence. "How would you approach an ambiguous project" produces a rehearsed answer that every candidate can give.

How do I evaluate a candidate's ability to work in ambiguity?

Give them an underspecified problem and watch whether they commit to an interpretation. Strong candidates state their assumption and proceed. Weaker ones ask clarifying questions indefinitely, which is correct behaviour in a structured company and a dependency you cannot support in an unstructured one.

How do I evaluate whether a candidate can grow into a larger role?

Look for scope that grew without a title change, and for evidence they built something that made their own previous work unnecessary. People who grow into larger roles tend to have expanded their remit before being granted it, rather than waiting for permission.

How do I hire managers who've never managed before?

Assess whether they already do the job informally: unblocking people, setting standards, giving hard feedback. First-time managers who succeed are usually already managing without the title. Be explicit that the job is different from the one they are good at, and check they actually want it.

How do I hire for future roles that don't exist yet?

Hire for the work in front of you and select, among people who can do it, for those who could plausibly grow into the larger version. Hiring purely for a future state buys capability you cannot use and often cannot keep, because the person is underemployed in the meantime.

How do I hire for team culture when I'm no longer interviewing everyone?

Replace your presence with a written standard and one trained interviewer per loop who owns it. Culture drifts when the criteria live in the founder’s head, because the people hiring in your place are guessing at what you would have decided.

Should I hire for culture fit or hire based on leverage?

Neither as usually framed. Culture fit tends to mean similarity, which narrows the team. Hire on values alignment plus demonstrated leverage: whether the person raises the output of those around them. That combination keeps the standard high without selecting for people who resemble the founders.

What's the right interview process for a scaling startup?

Four or five stages with distinct purposes and no repetition: a calibration call, a structured experience interview, a realistic work sample, a cross-functional conversation, and references that test unresolved claims. The failure mode is an unstructured chain of meet-the-team conversations that repeats questions and amplifies bias.

Should I hire a generalist or specialist at seed stage?

Generalist for anything still finding its shape, specialist where the work has stabilised and being wrong is expensive. At seed stage most work is still moving, so the default leans generalist, with narrow specialist hires where a specific technical risk justifies one.