How Long Does a Technical Interview Last? We Measured 403 Real Ones
Career guides put technical interviews at 45 to 60 minutes. Across 403 real interview sessions on GeekBye, the median ran 36.1 minutes, two in three passed half an hour, and one in seven passed an hour. Here is the whole distribution, what it means for your round, and what it cannot tell you.
A technical interview typically runs about 36 minutes. That is the median across 403 real interview sessions recorded on GeekBye between May 1 and September 8, 2026, by 112 different candidates. Two in three ran longer than half an hour. One in seven ran longer than an hour.
If you are booking time off, reading the calendar invite, or working out how much focus you need to hold, that is the number to plan around — with one asterisk we will get to: we are measuring how long a candidate's app was listening, not how long the interview was officially scheduled for.
The whole distribution, not just the average
Averages hide the part you actually care about, which is the tail. Here is the full shape of those 403 sessions.
| Measure | Value |
|---|---|
| Median length | 36.1 min |
| Ran longer than 30 minutes | 66% |
| Ran longer than 40 minutes | 41% |
| Ran longer than 60 minutes | 15% |
| Median, candidates on the free plan | 35.3 min |
| Median, candidates on a paid plan | 36.2 min |
Two details are worth pulling out.
There is no single "interview length" — there is a spread. A third of these rounds wrapped up inside half an hour. Two in five went past 40 minutes. Fifteen percent went past an hour. If you are preparing for one specific call, the honest answer is that anywhere between half an hour and an hour is ordinary, with a real tail beyond it, and your plan has to survive the whole range.
Free and paying candidates ran the same interviews. Medians of 35.3 and 36.2 minutes. That matters because it rules out the most obvious way this number could be skewed — that the people who pay for an interview assistant are the ones sitting the long, senior, high-stakes rounds. They are not. The length is the same on both sides.
Most recordings are short. Most interviews are not.
Widen the net and the picture inverts, which is worth understanding before you trust any app-derived number, including ours.
Across every ended session of 5 minutes or more between June 16 and September 14, 2026 — 913 sessions from 209 candidates — the median is 18 minutes, not 36. The quarter-mark is 9 minutes; three quarters of sessions finish inside 34.5 minutes; the 90th percentile is 56.5 minutes. Here is how they fall:
| Session length | Sessions | Share |
|---|---|---|
| 5 to 15 minutes | 397 | 43% |
| 15 to 30 minutes | 239 | 26% |
| 30 to 45 minutes | 127 | 14% |
| 45 to 60 minutes | 90 | 10% |
| 60 to 90 minutes | 32 | 4% |
| Longer than 90 | 28 | 3% |
Go wider still — all 3,748 sessions we have recorded, with no length floor at all — and the median collapses to 1.4 minutes. That is not a finding about interviews. It is a finding about software: most times someone presses a button in a new app, they are checking that it works.
This is why the headline number uses a 20-minute floor. Below that line we genuinely cannot separate a microphone test from a dropped call from an interview that finished early, and we are not going to pretend otherwise. What we can say is that 277 of those 913 sessions — 30% — ran half an hour or longer, which is a lot of long calls for a population where most recordings are short.
What 36 minutes means for the round you have on Thursday
Here is the part that changes how you prepare.
Most interview preparation is front-loaded. People rehearse the introduction, the two-minute summary of their background, the reason they are leaving. Then the measurement says this: in two thirds of these interviews, there is still time on the clock after minute 30. In two in five, after minute 40.
That long middle is the part that gets the least rehearsal. It is the part where the interviewer stops reading from a list and starts following up — the "why did you choose that" after you describe an architecture, the second and third question about the same incident, the extension to the coding problem you just solved. Preparation that covers the first five minutes brilliantly and thins out after twenty is preparation aimed at the wrong section of the call.
Two practical consequences.
Rehearse for endurance, not for the opening. If your mock sessions run 15 minutes, they are testing a different event than the one you are attending. Practice holding a technical thread for 35 minutes without your answers getting shorter and vaguer as you tire.
Do not book anything at the top of the next hour. One in seven of these rounds runs past 60 minutes. Scheduling the next commitment for the moment the invite says the interview ends means that in roughly 15% of cases you will be watching a clock during the most demanding part of it.
If you use a tool during the call rather than only before it, session length is also a budget question: our free plan includes 30 minutes of Listen audio a month, and the number on this page is the reason we are replacing that limit — a single median interview does not fit inside it.
What the published guides say
Our measurement is one input. It is worth setting against what the large career sites publish, because they say something slightly different, and the gap is informative rather than embarrassing.
Indeed's guide to interview length, published March 2, 2025, puts technical interviews at "45 to 60 minutes" and says they "may appear at any stage of the interview process." It puts in-person interviews at 45 to 90 minutes, phone screens at anywhere from 15 to 40 minutes depending on depth, and notes that "most job interviews last between 15 minutes and 90 minutes," with final rounds, panels and technical assessments sometimes running two hours or more.
So the published norm for a technical round sits above our median of 36.1 minutes. Both numbers can be right at the same time. A calendar invite for an hour is a booking. Our 36 minutes is the stretch a candidate actually had an app listening, and the two come apart in obvious ways: interviews start late, candidates press stop when the technical part is over, and the closing "any questions for us?" may never be recorded at all. Read our figure as the working part of the interview and the published 45-to-60 as the slot it sits inside, and they stop disagreeing.
What this data cannot tell you
The limits are as important as the numbers, so they get their own section rather than a footnote.
We measure listening time, not interview time. Every figure here is the duration of a recorded session. A candidate who starts the app three minutes into the call and stops it before the closing pleasantries produces a session shorter than the interview. We have no way to correct for that, and it almost certainly biases our numbers downward.
The 20-minute floor is a judgment call, not a fact. We use it as a proxy for "this was a real interview." It is a defensible proxy — 3,748 total sessions with a 1.4-minute median tells you what is down there — but a 12-minute recording could be a test, a dropped connection, a screen that ended early, or a real interview where someone forgot to press start. We cannot tell those apart, so we exclude them rather than describe them.
We cannot break this down by interview type. Sessions carry no label for phone screen, system design, pair programming, or on-site loop. Nothing in our data supports a per-type breakdown, so we have not published one. When we wanted per-type guidance above, we quoted a published source and named it. Everything in our own tables is undifferentiated.
This is one product's users. People who install an interview assistant are in an active search and skew toward software and technical roles. That is the population the numbers describe. It is not a random sample of everyone interviewing.
Method
Two populations, both aggregates over GeekBye Listen sessions, neither of which identifies an individual user.
The headline figures come from every Listen session between May 1 and September 8, 2026 that ran 20 minutes or longer: 403 sessions from 112 distinct users. Session length is the recorded duration from start to end. Percentages ("66% past 30 minutes") are the share of those 403 sessions above the stated threshold. The free-versus-paid split compares the plan the account was on at the time of the session.
The distribution table comes from a wider and more recent window: every ended session of 5 minutes or longer between June 16 and September 14, 2026 — 913 sessions from 209 users — bucketed by length, with the median, 25th, 75th and 90th percentiles computed over the same set. The two windows differ because they were pulled for different questions and at different times; they are not a like-for-like comparison, and the 20-minute-floor set is the one to quote for interview length.
The 3,748-session figure is every recorded session with no length filter, included only to show why a floor is necessary.
No figure on this page is modelled, extrapolated, or rounded up from a smaller number. Where we could not measure something — interview type, whether a short session was a real interview, how long the meeting was booked for — we said so instead of estimating it.
If your round is this week
Plan for 40 minutes and be unsurprised by 60. Rehearse the middle of the interview, not the introduction. And treat the published 45-to-60-minute guidance as the size of the slot rather than a promise about the conversation.
If you want the room to hold that middle stretch — real-time transcription of what the interviewer actually asked, and an answer while the question is still live — that is what GeekBye's Listen mode does during the call itself.
All figures are aggregates over GeekBye Listen sessions and never identify an individual user. Interview-length norms are quoted from Indeed, published March 2, 2025. If your technical rounds run consistently longer or shorter than what we measured, tell us — the sample gets better the more of it we can check against reality.
