How many interview rounds is too many?
About four. An internal Google analysis found that four interviews predicted a hiring decision with roughly 86% confidence, and that each additional interviewer past the fourth added about one percentage point. Structured interviewing genuinely works — that part is settled. What extra rounds mostly add is cost, paid in candidate-hours, and a filter on who can afford to keep showing up rather than on who can do the job.
Interview four predicts the hire. Interviews five through nine mostly predict who had the stamina to keep showing up.
First, the part that is genuinely settled
Structured interviewing works. When every candidate gets the same questions in the same order against the same scoring guide, the interview becomes one of the strongest predictors of job performance available — stronger than the unstructured conversation it replaced, and it predicts not just whether someone can do the work but whether they will contribute beyond the job description. If you have moved to structured interviews, you did the right thing and this article is not an argument against it.
What structure does not tell you is how many times to do it. And that is where processes have quietly inflated.
What Google found when it counted
Google put a number on this. An internal analysis by their staffing analyst Todd Carlisle, reported in Laszlo Bock’s Work Rules!, found that four interviews were enough to predict a hiring decision with about 86% confidence — and that each additional interviewer past the fourth added roughly one percentage point. Google cut its time-to-hire from around six months to about 45 days off the back of it.
Read that second number again. The fifth interviewer, the sixth, the panel you added last year because a hire went badly — each is buying about one point of confidence. They are not free. Somebody is paying for them, and it is not the company.
What validity does not measure
A selection method’s validity tells you how well it rank-orders the candidates who showed up and finished. It is a good number and it answers a narrow question.
It says nothing about who dropped out at round four because they took a faster offer. It says nothing about who never applied because the process looked like a second job. And it says nothing about someone who could do the work but could not show it, because every round was a performance and none of them looked like the job.
The two things a process is doing at once
Every hiring process is answering two separate questions. Only one of them tends to get tracked.
Questions in advance, a written bar that existed before the candidates did, the same structure for everyone, a chance to show work under conditions that resemble work.
Or a proxy for it — how someone discusses the work, under observation, on a clock, in a room that resembles nothing they will ever do again.
These are independent, and that is the finding worth carrying away. A process can be scrupulously fair and still be measuring the wrong thing — same questions, same order, same rubric, all pointed at a proxy. That is the quadrant that gets defended as rigorous, because everything visible about it is correct.
Three changes worth more than another round
- Count the candidate-hours. Multiply the people you interview per role by the hours each one spends — every interview, every take-home, every presentation, and the preparation each implies. That number is what your process costs the people going through it, and almost nobody has it.
- Cut to the shortest path that still sees real work. If a stage has never changed a decision, it is not assessment. It is attrition, and it is filtering on availability and stamina rather than ability.
- Move one interview question to a small real task. An hour, paid, resembling a normal Tuesday. Keep every bit of the structure you already built and point it at the work instead of at the conversation about the work.
The uncomfortable version
A long process feels like diligence. It reliably reads, from the inside, as taking the decision seriously. But past the point where prediction plateaus, additional rounds are not buying you a better decision. They are buying you the feeling of having made one — and the people who pay for that feeling are the candidates with the least ability to absorb the cost.
Read your own process
The Hiring Room Test asks eight questions about a hiring process — not about a candidate. It returns whether the process hands people a floor or makes them win one, whether it is assessing the job or a proxy for it, and what it costs in candidate-hours per hire, from your own numbers.
Take the Hiring Room Test →Sources. The four-interview finding is Todd Carlisle’s internal analysis at Google, reported in Laszlo Bock, Work Rules! (2015): four interviews predicted a hiring decision at about 86% confidence, with roughly one additional point of predictive power per interviewer thereafter. The characterisation of structured interviews as a leading predictor of job performance reflects the selection literature’s meta-analytic consensus. This page cites no other figures, because the ones circulating about interview length and drop-off mostly trace back to industry blogs quoting each other.
Common questions
How many interview rounds is too many?
The evidence points at about four. An internal Google analysis by Todd Carlisle, reported in Laszlo Bock's Work Rules!, found four interviews predicted a hiring decision with roughly 86% confidence, and that each additional interviewer past the fourth added about one percentage point. Past that, extra rounds buy very little prediction while continuing to cost candidate-hours — and they filter increasingly on who can afford to keep showing up rather than on who can do the work.
Do structured interviews actually work?
Yes, and this is not in serious dispute. Asking every candidate the same questions in the same order against the same scoring guide makes the interview one of the strongest available predictors of job performance, and it predicts contextual contribution as well as task performance. The open question is not whether to use structure. It is how many times to repeat it, and whether what you have structured is the job or a proxy for the job.
Can a hiring process be fair and still be wrong?
Yes, and that combination is unusually hard to spot, because everything visible about it is correct. Fairness and fidelity are independent. A process can give every candidate the same questions, the same order and the same rubric — genuinely fair — while every one of those questions asks how someone discusses the work rather than showing what they do with it. Everything visible about that process is correct, which is why it rarely gets examined.
What is the cheapest improvement to a hiring process?
Counting what it costs. Multiply candidates interviewed per role by the hours each one spends, including take-homes, presentations, travel and preparation. Almost no team has that number, and it usually changes the conversation immediately — because it reframes every extra round as a bill somebody else is paying, rather than as diligence the company is showing.
Related
- How job opportunities actually reach you — applied cold, referred, or sought out by name are three different kinds of access
- The Hiring Room Test — read your own process
- The audition gap — the same problem from the candidate’s side
- Why do the same people always talk in meetings?
- Why am I bad at interviews when I am good at my job?
- The founding paper, if you want the whole argument
A self-reflection tool based on the Taken or Given framework — not a validated psychological assessment or a diagnosis.
The four rooms, at work
The interview and the job are two different rooms, and people are two different performers in them. You get the four types and what each one needs from a workplace.