How to design an interview loop: focus areas, two reads on every must-have, order, timing and a loop audit

To design an interview loop, start from the decision it must make and the evidence still missing after the screens. Give each session two or three focus areas, make sure every must-have is assessed in depth by one session and checked independently by a second, choose the format that best shows each focus area, then order the sessions and fit them into a time budget the role justifies. Finish with independent scorecards and a debrief, and audit the loop after a handful of finalists to see which sessions are actually informing the decision.

A loop is the final block of interviews, not the whole process. The stages before it are designed in the interview process template, and the document you send interviewers is the panel interview plan template. This page covers the design decisions in between, with a worked loop and the arithmetic for auditing one.

What is an interview loop, and what must it decide?

Amazon uses the term in its own candidate guide, which says that in the loop "you'll meet with anywhere from two to seven Amazon employees", one of whom may be a Bar Raiser, "a specially trained interviewer" who acts as an objective third party (as of November 2026). The word is now used more widely for any final round of several back-to-back interviews, and this page uses it that way.

Whatever you call it, the loop has one job: to give the people deciding enough independent, job-related evidence to make an offer or not. Write that decision at the top of the design, along with three inputs:

• The scorecard, with must-haves marked. If there is none yet, build it first from an intake meeting.
• What earlier stages already settled. Logistics, pay fit and anything the hiring manager interview assessed in depth do not need retesting in the loop.
• The open questions, the things nobody has evidence on yet. These are the loop's real agenda.

How many focus areas should each session own?

Two or three. A 45-minute behavioral session has room for roughly two questions with proper follow-up probes; spread across six competencies, each gets a sentence of evidence. Data points the same way: in a Greenhouse analysis of about 10.3 million scorecards from jobs created in 2022, comparability between interviewers' scorecards was estimated to be 31 to 38 percent higher when an interview used assigned focus attributes, and the author suggested three to five per interview. For a loop, where several sessions share the load, the low end of that range is usually enough.

Why does every must-have need two independent reads?

One interviewer's rating of one answer is a noisy measurement. A second read from a different interviewer, using a different question and scored before any discussion, gives you a check you cannot get any other way. Schmidt and Zimmerman (2004, Journal of Applied Psychology) found that averaging three to four independent unstructured interviews gave about the same validity for predicting job performance as one structured interview by a single interviewer, which is a measure of how much a single loose read misses. How fast the gains from extra reads shrink is shown in how many interview rounds.

The rule that follows is simple to apply in a coverage map:

• P (primary): the session that assesses the competency in depth. Exactly one per competency.
• C (check): a second, independent read with a different question. Required for must-haves; optional for nice-to-haves.
• A competency with no P is a gap. A session with no P is a candidate for cutting.

Which format fits each focus area?

Take-home work belongs before the loop, not in it; the scoring rules are in how to evaluate a take-home assignment.

Who should sit in each session?

• Trained interviewers only. Someone who has not been through your interviewer training should shadow, not score; see interviewer shadowing.
• The same interviewers for every finalist, where feasible. OPM's structured interview guide (Appendix D) recommends using the same interviewers across all candidates to keep ratings consistent. Swapping an interviewer mid-search makes finalists hard to compare.
• One independent seat. Someone outside the hiring manager's reporting line, who has no stake in filling this role fast, in the spirit of Amazon's Bar Raiser.
• No two seats from the same angle on one must-have. The check read is worth more from a different vantage point, such as a peer and a cross-functional partner.

Who owns what on the day, from the coordinator to the debrief facilitator, is set out in interview panel roles and responsibilities.

How should you order the sessions and budget the time?

1. Open with a short, unscored welcome that explains the day, the sessions and the breaks.
2. Put the work sample early while the candidate is fresh, and its check read later in the day.
3. Leave the hiring manager's session for last if they also need to answer the candidate's questions about the team, so the candidate arrives with them.
4. Add a break after every two sessions, and never schedule a scored session over a meal.
5. Set a total budget before you fill it. Half a day of scored time is a lot to ask of someone in a job; justify anything longer by the role.

Worked example: a loop for a customer success manager

An invented role with invented numbers. The hiring manager interview has already covered renewal ownership in depth. Three must-haves (M) and three nice-to-haves (N) remain to cover:

COVERAGE MAP                 HM interview  S1 Escalation  S2 Behavioral  S3 Account   S4 Sales
                             (done)        role-play      (senior CSM)   review case  partner
Renewal ownership (M)        P             .              .              C            .
Handling escalations (M)     .             P              C              .            .
Product adoption (M)         .             .              P              C            .
Working with sales (N)       C             .              .              .            P
Using account data (N)       .             .              .              P            .
Written follow-up (N)        .             P              .              .            .

Focus areas per session:     S1: 2   S2: 2   S3: 3   S4: 1
Must-haves with two reads:   3 of 3

The candidate spends 3 hours 55 minutes, 3 hours 15 of them scored. Add 15 minutes per interviewer per session for the scorecard (six interviewer-sessions × 0.25 = 1.5 hours) and one loop costs about 6.9 interviewer hours. If, as an example, one finalist in three is hired, the loop alone costs about 21 interviewer hours per hire, which is the number to show anyone who wants to add a session.

What happens after the last session?

Every interviewer submits a scorecard before seeing anyone else's, then the debrief compares evidence one competency at a time. The reasons for the order are in independent feedback before the debrief, and the agenda is in the interview debrief template.

How do you audit a loop after it runs?

After ten or more finalists, check whether each session is earning its place. Two numbers per session, from the scorecards you already have:

Pass rate  = finalists the session recommended ÷ finalists who did the session
Agreement  = loops where the session's recommendation matched the final
             decision ÷ loops

Invented data for 12 finalists, 4 of them hired:

Read agreement with care: a session can agree with a poor decision, and a session that often disagrees may be the only one testing something the others miss. The audit tells you where to look, not what to cut. Two more checks complete it: how often the two reads on each must-have differ by two or more points (a sign the questions or anchors need calibration; see inter-rater reliability), and, a few months after the hires start, whether loop scores line up with quality of hire.

Common loop design mistakes