The bar: define it, write it down, hold it

Amazon's answer to this problem is a person. On every hiring loop there is a Bar Raiser, someone trained for the role, drawn from outside the hiring team, with a veto that the hiring manager cannot overrule. They have no stake in filling the seat. That is the entire design.

You can dislike a lot about how Amazon operates and still notice that this specific mechanism has been copied by almost every company that has had to hire quickly without falling apart. It works because it treats the bar as a structural problem rather than a matter of individual judgment.

Judgment is not the issue. Everybody's judgment is fine in January.

The bar is a document or it doesn't exist

Ask five interviewers at your company what a "yes" looks like for a senior engineer and you'll get five answers. They'll overlap by maybe sixty percent. That overlap is your real bar, and it moves.

A written bar is not a competency matrix and it's not the job description. It's a short, specific description of the person you would fight to keep. Two hundred words. What they can do without help. What they've done before. How they behave when the requirements turn out to be wrong. The kind of code review they leave.

Write it for one role, not for all roles. Write it by describing three people you'd hire again tomorrow, then removing everything that's true of only one of them.

Then put it at the top of the scorecard, so nobody can score a candidate without having just read it.

The three pressures

Bars don't collapse. They erode, and always through one of three routes.

The headcount target. Somebody committed to twelve engineers by Q3 in a board deck. That number now has more institutional weight than the quality of any individual hire, because the number is visible and the quality is not, at least not for six months. This is the most common cause and it's structural, not personal.

Exhaustion. The team has been short-staffed for four months. The people doing the interviews are the same people covering the gap. At some point a tired staff engineer looks at a borderline candidate and thinks: this person would take some of this off me. That thought is understandable and it is almost always wrong, because a borderline senior hire consumes more of that engineer's time in the first six months than the gap did.

Sunk cost. You've seen thirty-eight candidates. This one is the best of the thirty-eight. That fact carries no information about whether they're above the bar, but it feels like it does, and by candidate thirty-eight the whole loop wants it to be over.

Notice that none of the three is a failure of standards. They're all failures of design. Which is why exhortation doesn't fix them.

The fourth one, which isn't a pressure at all

Those three are situational. They arrive with a bad quarter and they leave with a good one. There's a fourth force that never leaves, and I'm going to put it in the first person because I don't think it's honest any other way.

I've hired too many people on vibes and on what makes me comfortable. Plenty of mistakes. And I'm the one saying raise the bar, out loud, to the whole org. I still lean back on comfort, because it's easier on my brain.

That's what it actually is. Comfort is a cognitive discount. Assessing an unfamiliar person properly is expensive, and somebody who feels familiar arrives pre-assessed, so you skip the expensive part without noticing that you skipped it.

Two tells, and I'd recognise both in myself now. The candidate reminds me of someone I'm close to. And the candidate doesn't push back on my philosophies.

The second is the dangerous one, because agreement is pleasant and it reads as judgment. What you're selecting for is a person who won't tell you things. That's the intake valve for the failure that makes bad news arrive late, installed at offer stage, months before anyone could observe it working. You don't need to react badly to bad news if you've hired people who won't bring you any.

It's also, I think, what "culture fit" is when you're the one deciding. The phrase gets attacked for producing homogeneity and the attack is right, but it lands on a label. The thing underneath the label is simpler and worse: recognising someone is cheap, and assessing them isn't.

Who holds the veto

The hiring manager should not have the final yes. They have the strongest incentive to fill the seat and the least ability to see the bar drifting, because they're the one under the pressure.

Three arrangements work. A dedicated bar raiser, on the Amazon model, if you have enough hiring volume to train them. A hiring committee that reviews written packets without the manager in the room, which is expensive but very consistent. Or, at small scale, a standing rule that one specific person outside the team must interview and can say no. At twenty engineers this is usually the CTO, and it stops being the CTO somewhere around sixty.

What doesn't work is consensus. Consensus on hiring converges on the least objectionable candidate, and least objectionable is a different thing from good.

There's a harder version of this that I've had to accept about myself. Criteria you enforce on yourself are advice. Criteria somebody else has standing to invoke against you are a gate. I override my own criteria. It's a very human thing to do, and I'd still tell you to write them, but writing them isn't the mechanism. Naming the person who is allowed to point at them and stop you is the mechanism, and then meaning it when they do.

Which is the real reason the veto has to sit with someone who isn't carrying the pressure. Including, and especially, when the person carrying the pressure is you.

The question underneath

Every rubric eventually reduces to one question, and it's worth asking it explicitly in the debrief, out loud, before anyone shares scores.

Would I want this person on the hardest problem we have?

Not "can they do the job." Not "are they better than the average person here," which is a formula for regression. For senior hires there's a stronger version: would I be comfortable reporting to this person in two years?

A "yes, but" is a no. The "but" is doing all the work in that sentence and it will still be there in eighteen months.

Debriefs

Two mechanics, both cheap, both routinely skipped.

Interviewers submit written feedback and a score before the debrief, not during it. Otherwise the most senior voice in the room sets the answer for everyone, and you've collected one opinion five times.

The debrief starts with the dissenting view. If four people said yes and one said no, the no goes first and gets uninterrupted time. Not because the no is right, but because a room that has already heard four yeses will not hear the fifth voice properly.

The part that's usually left out

Everything above assumes false negatives are cheap and false positives are expensive. At most companies, most of the time, that's correct: a bad hire costs somewhere between six months and two years of manager attention, team morale, and eventually a difficult exit.

It stops being correct in two situations, and I've seen both handled badly by people quoting the rule.

The first is when the empty seat has a real cost that nobody is pricing. If the team has been down two people for five months, the cost of the vacancy includes the burnout of the people covering it, which shows up later as resignations you don't connect back to the vacancy. A high bar plus a vacant seat is still a decision, and it should be an explicit one.

The second is early-stage hiring with no brand. At eight people, with no name, no comp premium, and a product that might not exist next year, a bar calibrated on what a well-known company can attract isn't a bar, it's a way of not hiring while feeling principled about it. The right adjustment isn't to lower the standard. It's to change what you're screening for: trajectory and range over polish, and a much shorter time to a real work sample.

The tell

You'll know the bar has moved when the sentence "for this role, at this stage, we should probably..." starts appearing in debriefs.

That sentence is never followed by a raise in standards.