---
title: How Interviewers Score Candidates: The Scorecard (2026)
description: How interviewers score candidates: the hire/no-hire call, the level pick and the 1–4 competency scores on a scorecard, and how to give evidence for each.
url: https://usegreenroom.app/blog/how-interviewers-score-candidates
last_updated: 2026-10-06
---

← Back to blog

Interview Prep

# How interviewers score you: the form they fill in after you hang up

October 6, 2026 · 13 min read

![How interviewers score candidates — the interview scorecard with a hire or no-hire call, a level pick and 1 to 4 competency scores, guide from Greenroom, the AI mock interviewer](/assets/blog/how-interviewers-score-candidates-hero.webp)

On September 15, a San Francisco startup called TypeSafe AI released **Jev**, an AI model with an odd selling point: it will not talk to you. It writes no essays and does no small talk. Jev answers exactly three kinds of question — yes or no, pick one from this list, or place this on a scale — and, by the company's own figures, most answers come back in about a tenth of a second. The launch video was reportedly watched around 40 million times on X, and new signups were paused within a week because of demand.

The tech world spent a fortnight arguing about whether a model that cannot chat is the future. Candidates should be paying attention for a different reason. The person who interviewed you last Tuesday has been doing the same job as Jev all along.

Picture it. You spent the evening rehearsing a two-minute story about the Kafka migration. You delivered it well, with a pause in the right place. The interviewer nodded twice. Then you hung up, and they opened a browser tab with a dropdown labelled *Overall recommendation*, a 1–4 scale next to the words *Problem solving*, and an empty box titled *Evidence*. Your Kafka story became one line in that box. On a bad day it became “talked about Kafka for a while.”

That form is the real subject of this guide: **how interviewers score candidates**, what the scorecard asks, what gets written in it, and how to give the interviewer something they can actually write down. One caveat first, and it matters. Greenroom has no insider access to any company's hiring panel, and **formats vary** by company, team and role. What follows is the common pattern, described publicly by companies like Google and by applicant-tracking vendors like Greenhouse, not any one employer's form.

## What an interview scorecard actually looks like

![Diagram of a typical interview scorecard — competencies chosen before the interview, a 1 to 4 score for each, an evidence box for each score, an overall call from strong no to strong yes, and sometimes a level pick such as SDE-1 or SDE-2](/assets/blog/how-interviewers-score-candidates-diagram.webp)

The common shape of an interview scorecard. Every row is a typed answer, and every typed answer needs a line of evidence behind it.

Most **interview scorecards** have the same five parts, whatever the software.

- **Competencies, chosen before you walk in.** Three to six per round, set by whoever designed the loop — problem solving, coding, communication, system design, ownership. Your interviewer is usually assigned one or two of them, not all.
- **A score for each.** Commonly a 1–4 scale. Four points is a deliberate choice: there is no neutral middle to hide in, so every interviewer has to lean one way.
- **An evidence box for each score.** What you said or did that justifies the number. Interviewers are usually trained to write behaviour, not adjectives.
- **The overall call.** Greenhouse's default, for example, asks for one of four options, from “Definitely not” to “Strong yes.” Other tools use **hire / no hire / strong hire** wording for the same thing.
- **Sometimes a level pick.** SDE-1 or SDE-2, L4 or L5. This is where an offer quietly becomes a down-level.

## Yes or no, pick one, or a score: the three shapes of every verdict

Look at that list again and you will see Jev's three question types. The overall call is a yes or no with a confidence attached. The level is a pick-one. Each competency is a score on a scale.

That is not a coincidence. Jev is built that way because software needs typed answers it can act on, and “the vibe was pretty good” is not something a program can branch on. Hiring processes moved to scorecards for a similar reason: five interviewers' free-text impressions cannot be compared, and five sets of 1–4 scores can.

There is research behind it too. A widely cited 1998 meta-analysis by Frank Schmidt and John Hunter found structured interviews predicted job performance noticeably better than unstructured ones, and Google's re:Work guidance on structured interviewing recommends the same thing in practice — fixed questions, defined rating scales and written evidence. Treat the exact effect sizes as directional, but the direction has held up for decades.

**The mechanism:** your interviewer is not deciding whether they liked you. They are filling in a form that asks for evidence, and you can only be scored on what you said out loud, in a shape they could write down.

## How interviewers score candidates on each competency

Each 1–4 score is usually tied to written anchors, so two interviewers mean roughly the same thing by a 3. Exact wording differs everywhere, but a problem-solving anchor set often looks something like this:

- **1 —** could not make meaningful progress, even with hints.
- **2 —** reached a working approach, but only after significant hints.
- **3 —** solved it, explained the tradeoffs, needed at most a small nudge.
- **4 —** solved it cleanly, drove the discussion, and raised edge cases before being asked.

Notice the word that keeps appearing: **hints**. In most rubrics the help you received is recorded, not forgotten. The interviewer who rescued you after four minutes of silence was being kind in the room and accurate on the form. That is why a friendly interview can still produce a 2.

## What interviewers write in the feedback

The evidence box is where most decisions are actually made, because it is what the debrief reads. Good interviewers are trained to write what happened rather than how it felt. Compare two sets of notes on the same candidate:

> “Seemed smart. Good communication. Some gaps on scaling.”

> “Proposed a hash map without prompting, stated O(n) time and space. Missed the empty-input case until asked. On scaling to 10x traffic, suggested caching but could not say what to invalidate.”

The second set is fairer to you, and it is also more dangerous, because it can only contain what you said. If your reasoning happened silently in your head, the box records the silence.

So the practical test for any answer is: **could the interviewer copy a sentence of it straight into the evidence box?** “We improved performance a lot” cannot be copied. “I moved the report query off the primary database and cut p95 from 1.4 seconds to 380 milliseconds” can, almost word for word.

## Fast first impressions vs the slow form

Jev's makers call it a “System One Model,” borrowing Daniel Kahneman's terms from *Thinking, Fast and Slow*: System 1 is the fast, automatic judgement, System 2 the slow, deliberate one. Interviewers have both. Within the first couple of minutes, most people have formed an impression, whether they want to or not.

The scorecard exists to force System 2. It makes the interviewer justify a number with evidence, often before they see anyone else's scores. It does not fully work — first impressions leak into scores, and every recruiter knows it — but it shifts the weight. A clear, calm opening still helps. The middle forty minutes, where evidence is produced, matter more.

## Hire, no hire, strong hire: what happens in the debrief

Once every interviewer has submitted, the scorecards meet. At many companies scores are submitted independently first, so nobody anchors on the loudest person in the room. Then comes a debrief or a hiring committee — Google's committee review and Amazon's bar raiser are the best-known versions, and both have been described publicly in broad strokes.

A few patterns are common enough to plan around:

- **One strong no usually outweighs two mild yeses.** A “lean no” with specific evidence is hard to argue past.
- **Mixed loops sometimes buy an extra round** rather than a rejection, particularly when the disagreement is about level rather than ability.
- **The evidence boxes get read aloud or quoted.** A strong yes with an empty box carries less weight than a yes with three concrete lines.
- **Timing is slower than it feels.** Scorecards are often due within a day or two, but debriefs, approvals and level discussions commonly stretch a decision to one or two weeks.

If the silence after a final round is getting long, our guides on [signs an interview went well](/blog/signs-an-interview-went-well-or-badly) and on being [ghosted after a final round](/blog/ghosted-after-final-round-interview) cover what the wait does and does not mean.

## How to give the interviewer something to write down

None of this means performing for a form. It means making the evidence you already have visible.

- **Think out loud.** Silent progress cannot be scored. Say what you are considering and why you are dropping it.
- **Name the tradeoff yourself.** “The cost of this approach is memory” is a sentence that goes straight into the problem-solving box.
- **Use numbers and say “I.”** Your decision, your metric, your result. “We” is a team the interviewer is not hiring.
- **Close the loop.** End a technical answer with one summary line: “So — hash map, O(n), and it handles empty input.” You have just written the evidence box for them.
- **Ask for a hint deliberately.** A hint requested early and used well reads very differently from one extracted after four minutes of silence.
- **Use the last five minutes.** A good question can add evidence too, as our [questions to ask the interviewer](/blog/questions-to-ask-the-interviewer) guide explains.

For how each round in a loop sets its own competencies, see [the interview process, every stage explained](/blog/interview-process-stages-explained) and our guide on [how to prepare for a final round interview](/blog/how-to-prepare-for-a-final-round-interview).

## Where practice tools help, and where they don't

Most prep options train one row of the scorecard and leave the rest blank.

**Glassdoor and GeeksforGeeks interview experiences.** Good for knowing which questions come up. They tell you nothing about how the answers were scored, and the writers rarely saw their own scorecard.

**LeetCode.** Excellent for getting to a correct answer. It trains you to solve silently, and it marks correctness only — one row out of four or five.

**Peer mocks, Pramp-style.** A real human asking real questions. But peers are polite, and very few will write “needed significant hints” about someone they will match with again.

**ChatGPT.** Useful for drafting answers, but it tends to be agreeable and generous unless you push it hard. Ironically, its biggest strength — it will chat about anything — is the opposite of how the form works.

**Paid human mocks with experienced interviewers** (interviewing.io and similar). The closest thing to a real scorecard, from people who have filled in thousands. The tradeoff is cost per session.

**Greenroom.** I built Ari, the AI interviewer behind Greenroom, after years of walking out of interviews sure I had explained myself well, and later realising the interviewer had heard a vaguer version of me than the one in my head. Ari asks the follow-up instead of nodding along, does not rescue you after a long silence, and scores you 1–10 across five dimensions with quotes from your own transcript — the closest thing to reading your own evidence box. Our [feedback report explainer](/blog/ai-interview-feedback-report-explained) shows exactly how. Honest tradeoff: Ari's score is not your target company's scorecard. It cannot reproduce a specific team's calibration or the politics of a debrief, and the practice is spoken only, with no live code editor.

## The one-line version

Every interview ends as a yes or no, a pick-one and a handful of 1–4 scores, each justified by a sentence the interviewer wrote about you. Jev answers in a tenth of a second. Your interviewer takes twenty minutes. Give them sentences worth copying down.

## Frequently asked questions

### How do interviewers score candidates?

Most companies use a scorecard. Each interviewer is assigned a few competencies, such as problem solving or communication, and scores each one, commonly on a 1 to 4 scale tied to written anchors. Every score is backed by an evidence note describing what the candidate said or did. The interviewer then gives an overall recommendation, usually on a scale from strong no to strong yes, and sometimes picks a level. The scorecards are compared in a debrief or hiring committee, where the final decision is made.

### What is an interview scorecard?

An interview scorecard is the structured form an interviewer fills in after a round. It lists the competencies that round is meant to assess, a rating for each, a box for evidence supporting each rating, and an overall hire or no-hire recommendation. Its purpose is to make different interviewers' judgements comparable and to push them towards evidence rather than gut feeling. The exact fields vary by company and by the applicant-tracking tool it uses.

### What do interviewers write in interview feedback?

Well-trained interviewers write observed behaviour rather than adjectives, for example that a candidate proposed a particular approach, stated its complexity, missed an edge case until prompted, or needed a hint to progress. They note how much help was given, because most rubrics record hints. Vague notes such as seemed smart carry little weight in a debrief, which is why specific, quotable answers from the candidate tend to produce stronger feedback.

### What does strong hire mean in an interview?

Strong hire, or strong yes, is the top option on most overall-recommendation scales. It signals that the interviewer is confident and would argue for the candidate in the debrief, usually because they saw clear evidence above the bar on the competencies they assessed. A plain hire or yes means the candidate met the bar. Strong ratings carry more weight when the evidence notes behind them are specific.

### How long after an interview do they decide?

Interviewers are often expected to submit their scorecards within a day or two, but the hiring decision usually takes longer because every interviewer's feedback has to be collected, discussed in a debrief and sometimes approved by a hiring committee or a senior manager. One to two weeks after a final round is common, and level or compensation discussions can extend that. A long wait on its own is not a reliable sign of rejection.

### Does one bad interview round mean rejection?

Not always. Debriefs weigh the whole loop, and a weaker round can be outweighed when the other rounds show strong evidence, particularly if the weak round was outside the role's core skills. A strong no with specific evidence is harder to overcome than a mild no, and mixed results sometimes lead to an extra round or a different level rather than a rejection. It varies by company, so the safest assumption is that every round counts.

Your interviewer can only score what they can write down. [Greenroom](https://usegreenroom.app/) lets you practise out loud with Ari, who asks the follow-up instead of nodding along and shows you the evidence behind your score. Free to start. See [how AI mock interviews work](/blog/ai-mock-interview).