Philosophy & the Bar
The bar you tolerate is the bar you have. • A hiring system and a performance system are the same instrument, pointed at two different moments in time.
Overview
Most teams treat hiring and firing as unrelated events run by different people with different vocabularies. They are not. They are the two ends of one system: a written definition of what "good" looks like, applied when someone joins and applied again while they stay. If the two definitions disagree, you will hire people you later manage out for failing a bar they were never actually screened against. This system keeps a single yardstick from the first recruiter call to the last day.
This first part sets the beliefs the rest of the system depends on. Everything downstream — the loop, the scorecard, the improvement plan, the exit — is just these principles made mechanical.
Hire and fire against one written bar
A decision made from memory drifts. A decision made from a written rubric is auditable, teachable, and defensible. The single most valuable artifact in this entire system is one page that lists the pillars a person is evaluated on, what strong looks like per pillar, and how those pillars map to levels. That page is referenced by the recruiter screen, by every interview, by onboarding, and by every performance review.
When the hiring bar and the performance bar are literally the same list of pillars, three good things happen: interviewers and managers argue about the same thing, a new hire knows on day one exactly what they will be measured on, and "underperformance" becomes a specific, evidenced gap against a known standard rather than a feeling.
Concreteness beats sophistication
The single most reliable signal, on both sides of the system, is a named moment: scope, stakes, what was decided, the outcome, and a number. A candidate who says "I care deeply about reliability" is describing an aspiration. A candidate who says "I cut a service's p99 from 900ms to 210ms by moving connection pooling in front of the database, which held during our Black-Friday load" is describing a fact you can probe.
Reward the concrete and down-score the generic — the textbook answer, the vendor phrasebook, the "we decided" that hides whether the person decided anything. The same rule governs performance conversations: "communicate better" is not feedback; "flag blockers within a day instead of going quiet for a week" is.
Two cultural tells, weighted heavily
Two behaviors predict success far out of proportion to their size, so they are watched for explicitly:
First, "I don't know" once is fine; bluffing once is not. Admitting the edge of your knowledge is a strength; manufacturing confidence over a gap is disqualifying, because a person who bluffs in an interview will bluff in an incident.
Second, disagreeing with the interviewer, with evidence, scores up. A candidate who says "I wouldn't build it that way, and here's why" and is specific reveals more than one who recites best practices. The interview is not a one-way test; it is a simulation of working together.
| Do | Don't |
|---|---|
| Bring concrete moments — scope, stakes, outcome, numbers | Generic, textbook, or vendor-phrasebook answers |
| Say "I don't know" once when you don't | Bluff — once is enough to count against you |
| Push back with evidence when you think we're wrong | Treat the conversation as a one-way exam |
| Show how you actually work day to day | Describe how work is supposed to be done in theory |
Manage out with dignity, and never by surprise
The firing half of the system rests on two commitments. Manage out with dignity: the goal is never punishment, it is resolving a mismatch between a person and a role for the sake of customers, the team, and the person's own next chapter. Every message and meeting should be one you would be comfortable having read back to you.
No surprises: by the time anyone reaches a termination conversation, the outcome should already be understood by the person. If someone is genuinely blindsided at the exit, the manager failed at feedback — not the employee at performance. That single commitment is what forces continuous 1:1 feedback, a rubric the person has already seen, and an explicit improvement window before any decision.
Speed where it's cheap, fairness where it counts
Optimize the machinery for speed: a lightweight twice-a-year review, self-contained interview rounds, fast debriefs. Never optimize an individual's due process for speed: written notice, a fair improvement window, correct final pay, and legal/people sign-off are not places to cut corners. "Better good than perfect" is how you build the system; it is not how you end someone's employment.
The rest of this system is these six ideas turned into steps.