The debrief, level expectations and your four-week plan

The debrief converts five opinions into one decision at one level. Knowing how the arithmetic works tells you which failures are recoverable.

It also explains the second decision candidates forget about: the level. The same story earns different scores at different levels, and that is the most common reason a technically strong candidate gets downlevelled. This lesson covers both, then turns them into a four-week preparation plan.

How the scores combine

Most companies use a four- or five-point scale per interviewer, roughly: strong no hire, no hire, hire, strong hire. The scores are not averaged. They are argued.

Three rules govern nearly every debrief:

1. Evidence beats assertion. An interviewer who says "I just didn't get a good feeling" loses to one who can quote what you said. If your weakest round produced a vague negative and your strongest produced specific positives, you are in better shape than the raw scores suggest.

2. Independent agreement is decisive. Two interviewers who separately recorded the same concern have found a trait. Two who disagree have found a bad round.

3. Unfilled rubric rows count against you at senior levels and are neutral at mid. "No evidence of strategic thinking" is fine for a mid-level candidate and fatal for a staff one.

What "no hire on leadership, hire on technical" resolves to

This is the most common split, and it resolves in one of four ways:

The fourth row is worth reading twice. Concerns about capability are negotiable — the company can decide the gap is coachable. Concerns about how you treat people are treated as a fixed property, and no amount of technical strength offsets them.

The hiring committee

Some large companies add a committee that never met you and reads only the written packet: the write-ups, your resume, and the recruiter's notes.

This has one direct consequence for you: anything an interviewer did not write down does not exist. A brilliant point you made in passing, which the interviewer enjoyed but did not record, is invisible to the people making the decision. This is the strongest possible argument for the signposting in How interviewers are trained and how they score.

What the level decision hangs on

Level is decided by scope, autonomy, and influence in your stories — Level as the hidden variable develops this triad in full. In debrief terms:

• Scope: how big was the thing you affected? One service, one team, one org?
• Autonomy: who decided? You, your lead, or your manager?
• Influence: how many people did you move who did not report to you?

A candidate whose stories are technically excellent but always scoped to their own tasks gets "hire, mid level" — even if they applied for senior.

Level expectations at a glance

The mismatch that costs offers

A senior candidate tells this story. The API in it is their product's application programming interface — the endpoint other services call for product data.

"The ticket asked for a caching layer on the product API. I looked at the access patterns, found 80% of the traffic hit twelve products, and used a small in-memory cache instead of the Redis cluster the ticket specified. Latency went from 900 ms to 40 ms."

That is a clean story and it is scoped mid. The candidate was given the what and chose a better how. Nobody outside the ticket was moved.

The senior version of the same work:

"The ticket asked for caching on the product API. Before building it I checked whether the problem was actually cache-shaped — 80% of traffic hit twelve products, so I proposed a small in-memory cache rather than the Redis cluster in the plan, which saved us the operational cost of another cluster. I wrote that up, took it to the platform team who owned the Redis budget, and they adopted the same pattern for two other services. Latency 900 ms to 40 ms, and we avoided a cluster we'd have paid for indefinitely."

Same engineering. The second version shows the decision, the trade-off, and two other teams following. That is what moves the score.

Your preparation plan

Knowing how the decision is made tells you what to prepare; the plan tells you when. Four weeks, about five hours a week. It is time-boxed because most people doing this have a full-time job and an interview loop already scheduled. (If your loop is days away, use the one-week path from the introduction instead.)

Notice: no week is more than about five hours, and the mock is in week 4, not week 1. A mock before you have stories produces feedback about material you were going to change anyway.

What each week produces

Week 1 — inventory. A list of twenty candidate projects and a filled coverage matrix (Mapping yourself to the rubric) showing which competencies you have nothing for. The output is a gap list, and it is usually uncomfortable. Most engineers find two or three competencies with no material at all.

Week 2 — drafting. Six story cards in the story-card format. Written, not spoken. Expect each first draft to run six minutes when spoken and need cutting to three and a half.

Week 3 — drilling. The competency sections for your level, plus prepared answers to the follow-ups. Then record four stories on your phone and listen back. This is the least pleasant hour of the four weeks and the highest-yield one.

Week 4 — mock and repair. One full mock against the scoring sheet from Running a useful mock, then repair. Do not write new stories in week 4.

Where the time actually goes

Candidates habitually spend 80% on reading and writing and 5% on speaking. That ratio is why prepared candidates still ramble in the room.