How to design an interview loop

Design it backwards in five steps

1. Write the three or four requirements the role genuinely has.
2. Assign each requirement to exactly one stage that can observe it.
3. Add a stage only if a requirement has nowhere to live.
4. Remove any stage that assesses nothing unique.
5. Write the coverage map down and give it to every interviewer.

A coverage map, not a schedule

Most loops are designed as a sequence of conversations with people who are available, and the coverage emerges by accident. Invert it. Write the requirements in a column, the stages in a row, and put exactly one mark in each requirement's row. The usual result is uncomfortable: three stages all assessing problem solving, nothing assessing whether the person can work without supervision, and a round that exists because a particular team lead has always done it.

A four-stage loop that covers most engineering roles

• Screen (30 min) — relevant experience, communication, motivation, logistics. Owns nothing else. See phone screen questions to ask candidates.
• Technical assessment (60 min) — problem decomposition, verification, code readability. See what to look for in a coding interview.
• Craft or design discussion (60 min) — trade-off reasoning, handling ambiguity, depth in the domain.
• Hiring manager (45 min) — ownership, collaboration, motivation and level calibration.

Say what each stage does not own

Writing down what a stage is not responsible for is as valuable as writing down what it is. Interviewers who have not been told otherwise will rate everything they glimpsed, which produces four correlated opinions about communication and no evidence about anything else. The instruction “you are not assessing technical depth in this round” is one line and changes what gets recorded.

How many stages

• Three to four for most roles. Two is usually too few to see more than one dimension; five starts costing you candidates.
• Count total hours, not stages. Four 45-minute conversations is a smaller ask than three two-hour ones plus a take-home.
• Count the candidate's calendar, not yours. Four stages across four separate weeks is a six-week process, and six weeks is where good candidates accept something else.
• Onsite blocks are one stage for you and one long day for them. Build in breaks and say when they are.

Sequence to fail fast and fairly

Put the stage with the highest rejection rate early, provided it is a fair filter. This respects the candidate's time more than a gentle screen followed by a brutal fourth round. The exception worth making: do not put the most artificial stage first, because the candidates most likely to fail it early are often the ones who would do the job best — they are simply out of practice at being assessed.

Decide the bar before you start, not after

Write down what a hire looks like before the first candidate. Loops without a pre-agreed bar drift toward whoever has been seen most recently, and after six candidates the bar reflects the pool rather than the role. If the pool is genuinely weak, that is a sourcing problem, and lowering the bar silently converts it into a performance problem a year later.

Who should be in the loop

• The hiring manager, always, and ideally not last — they own the decision and should hear the candidate early enough to steer the remaining rounds.
• At least one person who will work with the hire daily. Peer interviewers catch things managers miss, and candidates ask them more honest questions.
• One person from outside the immediate team for senior roles, as a check on a team hiring entirely in its own image.
• Not everyone who wants to be. A loop is not a consultation exercise. Every additional interviewer adds scheduling latency and correlated opinions.

Brief your interviewers properly

An interviewer who has not been told what they own will assess general impressiveness. Send three things before each round: the attributes this stage owns, the attributes it does not, and the questions to ask. This takes one message and is the difference between four structured assessments and four conversations. New interviewers should shadow one loop and be shadowed on one before running a round alone — see how to calibrate interviewers.

Review the loop against outcomes

Twice a year, look at which stage rejected the people you later wished you had hired, and which stage passed the hires who did not work out. If a stage has never changed a decision, it is a scheduling cost with no information value. See how many interview rounds is too many.

Where Interviewer Mode fits

Disclosure: we make CoPilot Interview, which includes Interviewer Mode for the hiring side and a real-time assistant used by candidates. We have an interest in you using the first, and our view of the second is that interviews should be designed to measure something an assistant cannot fake. Weigh the advice below accordingly.

Frequently asked questions

How many interview stages should there be?

Three or four for most roles. Design backwards from requirements: assign each requirement to exactly one stage and add a stage only when something has nowhere to live. Count total candidate hours and elapsed weeks rather than the number of conversations.

What should each interview stage assess?

Exactly one set of attributes that no other stage owns. A typical engineering loop gives the screen relevant experience and motivation, the technical round decomposition and verification, the design round trade-off reasoning, and the hiring manager round ownership and level calibration.

What order should interview stages go in?

Put the highest-rejection fair filter early so you do not spend candidates' time, with one exception: avoid leading with the most artificial stage, since out-of-practice candidates who would do the job well fail those disproportionately.

How do I know if my interview loop is too long?

Check for duplicate coverage and elapsed time. If two stages assess the same attribute, one is redundant. If the loop takes more than about three weeks end to end, you are losing candidates to faster processes regardless of how good your stages are.