In 174,558 interviews you can line the booked length up against what actually happened. 54.3% of them ran past the slot, and among the 94,823 that did, the median overrun was 11.9 minutes.¹
The reflex reading is that interviews are too long. It's worth slowing down there. An overrun is a disagreement between a calendar entry and a conversation, and the number on its own can't tell you which of the two was wrong.
The rest of this piece is about what you can do with that disagreement. It isn't a case for cutting every stage to 30 minutes, and the data behind it wouldn't support one.
What an overrun tells you
A stage that runs long tells you the plan and the conversation disagreed. That's all it tells you. Maybe the slot was too short for the work. Maybe the stage got asked to cover more ground than fits, or the interviewer took up room the candidate needed. All three produce the same overrun.
The stage is the unit worth watching. One conversation running fifteen minutes long is a person being interesting. The same stage running long across a month of candidates is a scoping problem, and you can fix a scoping problem.
Most of those overruns were material. More than half of interviews ran over at all, and 44.1% ran over by five minutes or more. Set those two against each other and roughly one overrun in five was small enough to be calendar drift. The rest were long enough that somebody's afternoon moved.
For more on what good interviewing looks like, read our guide to what separates a good interviewer from a weak one.
Length follows the stage
There's no single right length, and the spread by stage is wide enough to make any universal target look arbitrary. Technical interviews average 62.6 minutes. Intake calls come in at 35.3, debriefs at 30.5, and across the corpus as a whole the median interview runs 35.5 minutes.¹
The overall direction is downward. The median length of interview captured on Metaview fell from 38.6 minutes in Q2 2024 to 32.6 minutes in Q2 2026, and 35.0% of interviews in the corpus run 30 minutes or less.¹ The trend is real, but the reason for it isn't in the data. That same fall is consistent with better-scoped stages, with a changing mix of stages, and with schedulers defaulting to shorter calendar blocks.
Source: Metaview interview data, 5.5M captured conversations
So the real question about a 30-minute screen is a scoping question. What's that stage supposed to establish, and does 30 minutes get you there? A recruiter screen and a case discussion answer that differently, and a process that gives them the same block has decided the answer by accident.
Where the obvious conclusion breaks
The tidy version of this argument would be that shorter interviews are tighter interviews. Two cuts of the same data get in the way.
Interviews that follow a loaded guide run longer. They average 39.5 minutes against 34.7 for unstructured ones, and only 19.1% of interviews in that sample work from a guide at all.¹ Whatever discipline a written guide adds to a conversation, it doesn't arrive as a shorter conversation.
Advance rates rise with length. Candidates in interviews of 30 minutes or less advanced 14.2% of the time. In the 31 to 60 minute band it was 21.5%, and in the 61 to 90 band it was 30.4%.¹ That's an association. It doesn't show that longer interviews work better, and the obvious confounder is stage: long interviews are disproportionately late-stage interviews, and late-stage candidates have already been filtered once. The narrow point stands anyway. If you came to this corpus looking for evidence that short is better, it isn't here.
There's a harder boundary underneath both of those. This record covers what happens inside the hiring process, and it stops at the offer. Nothing in it describes how anyone performed once they started. So no figure on this page, including the ones we like, can be read as a statement about quality of hire. Anyone whose interview data settles that question is describing data they don't have.
Hiring well takes time. But too many companies treat it like a box to check or worse, a burden to minimize. Companies that treat hiring as a cost try to minimize the time it takes. Companies that treat it as an investment build systems and expectations around doing it well.”
That's the honest shape of the problem. Time spent interviewing isn't automatically waste, and length by itself tells you very little. Hunt for the stage whose job nobody can describe.
What to do with a stage that keeps running over
Write down what the stage exists to establish before you touch the calendar. A small set of competencies, the questions that test them, and an agreement on what strong evidence sounds like. Once you've written down the job of a stage, an overrun becomes readable: either the job is bigger than the slot, or the conversation wandered off the job.
Then check three things, in this order:
- Scope against slot. Hold the job of the stage up against the time you gave it. A technical exercise scheduled like a screen will run over every time, so change the calendar.
- Duplication across stages. If two stages collect the same evidence, one of them is the candidate telling the same story twice. That's the cheapest time in the process to recover.
- The talk split, before you blame the candidate. The median interviewer takes 46.8% of the talk time, and only 5.6% of interviews are interviewer-dominated above 70%, across roughly 2,657 qualifying interviews with speaker-attributed audio.¹ Interviewer monologues are real and worth catching. The distribution says they're not the usual reason a stage runs long.
Then set a default that a stage is allowed to break. Thirty minutes is a reasonable starting block for many screens. Technical exercises, case discussions, and working sessions need longer, and saying so in advance costs less than finding out live.
What the record can and cannot tell you
Metaview Reports lets a team query its own interview data, and the structured notes from each conversation keep the evidence somewhere you can go back to. You end up with a record where you used to have a recollection. The reading is still yours: which stage keeps running over, whether two stages are asking for the same thing, and whether the job you wrote down for a stage is the job it's doing.
None of that is a quality score. It tells you where to look, and a human still has to look.
Candidates are running a clock while you work this out. Metaview's 2026 AI & Hiring Alignment Report found that 67% of teams lose qualified candidates to competitors who move faster, every month.¹ That's a survey of how teams experience their own speed, and it says nothing about interview length. It does mean a stage nobody can explain is an expensive thing to keep.
For a sense of scale, one customer-reported result: Sendwave reports a 28% reduction in interviews per hire, which its case study attributes to ongoing interviewer coaching. It's one company's before-and-after, with no controlled comparison behind it, so treat it as one example.
Start with one role. Map its stages, write down the job of each, pull a few weeks of interviews, and find the stage that can't explain itself. For the writing-it-down part, our guide to an impactful interview scorecard is the practical companion.
Look at what your interviews already contain.
Metaview captures the conversation, so you review what was actually said.
Frequently asked questions
Does an interview running over mean it was too long?
Not on its own. An overrun means the scheduled slot and the conversation disagreed, and the number doesn't say which one was wrong. A single long interview is a person being interesting. The same stage running long across a month of candidates is a scoping problem.
How long should an interview be?
Long enough to do the job that stage was given. In the corpus, technical interviews average 62.6 minutes, intake calls 35.3, and debriefs 30.5. A single default applied to every stage ignores the work each one does.
Are shorter interviews tighter interviews?
Not reliably. Interviews that follow a loaded guide run longer than unstructured ones, 39.5 minutes on average against 34.7. Whatever discipline a written guide adds, it doesn't show up as saved time.
Does shorter interviewing improve quality of hire?
This data doesn't answer that. The record covers what happens inside the hiring process and stops at the offer, so there's nothing in it about how anyone performed once they started. Interview length describes the process, and the outcome sits outside what this record can see.
Where do these interview figures come from?
Length, overrun, and stage figures come from Metaview's corpus of 5.5 million captured conversations. Talk-time figures come from a much smaller slice, roughly 2,657 interviews with usable speaker-attributed audio. Survey figures come from the 2026 AI & Hiring Alignment Report, based on 505 recruiting leaders and hiring managers in North America and EMEA.
¹ Source: Metaview's corpus of 5.5 million captured conversations, 2026. Interview length, overrun, and stage figures are drawn from platform data. Talk-time figures come from a smaller sample of interviews with usable speaker-attributed audio. Survey figures are from the 2026 AI & Hiring Alignment Report, based on 505 recruiting leaders and hiring managers in North America and EMEA.
