What an interview costs

If you've worked as an engineer at a big tech company, you've probably been pulled into an interview loop. You get a calendar invite, skim the resume ten minutes before, ask your favorite question, write feedback that evening and argue about the candidate in a debrief two days later. Most engineers never add up what that costs. The math is quick.

Say a team brings five candidates onsite to fill one role, and each candidate meets four or five engineers. That's over 20 hours of interviews. Each hour on the calendar costs about two^[1], because the interviewer also prepares, writes feedback, sits in the debrief and then needs time to get back into their own work. Add the phone screens before the onsite and the offers candidates turn down, and one accepted hire costs about 100 engineer-hours^[2]. A senior engineer at Google makes around $420,000 a year in total compensation^[3], so that's about $20,000 of engineering time for one hire, before you count what those engineers didn't build that week.

Most of that cost comes from how many people you interview, more than from how long each interview runs. Most of those hours go to candidates who don't get the job.

The cost of technical interviews

The loop above is a heavy one, and the rest of the industry isn't far behind. Across more than 100 million applications, the average technical hire now takes 23.3 hours of interviews, nearly twice a business hire^[4].

Those are calendar hours. They leave out the prep, the write-up, the debrief and the context switch. A programmer takes 10 to 15 minutes to start editing code again after an interruption^[5], and an interview is a much bigger interruption than a Slack message. When we added that overhead stage by stage, a typical company spends about 70 engineer-hours, or about $9,000 of engineering time, per technical hire^[2]. Companies that run technical interviews for other companies put it at 40 to 100 hours^[6].

Most of that time goes to people who aren't hired. In the loop above, four of the five onsite candidates leave without an offer, and so does everyone screened out before them. Multiply it out and US companies spend roughly $3 billion a year of engineers' time interviewing other engineers, and about $14 billion worldwide^[2].

You may have seen a bigger number. The widely quoted 60 billion dollar figure comes from a 2021 vendor funding announcement. It rests on a guess of 80 or more hours per hire and never shows the rest of the math^[7]. Our bottom-up estimate puts the global figure at about a quarter of that, and it doesn't need inflating.

What it costs candidates

Now look at it from the other chair. An engineer looking for a new job is usually still doing their current one. They spend evenings and weekends relearning algorithms they haven't touched since school, about six and a half hours a week on average^[8]. Then they go through the same kind of loop at several companies, each one starting from zero, each with a different set of engineers asking their own favorite questions. Some of them get a take-home that eats a Saturday.

Candidates notice when a process wastes their time, and some walk away. One in five job seekers say they have turned down an offer because of a bad interview experience^[9]. A loop that cost the employer 100 engineer-hours can still end with the candidate saying no.

Structure does the work

None of this means interviews don't work. A 2022 re-analysis of decades of hiring research put structured interviews at the top of every predictor of job performance it measured, more than twice as predictive as unstructured ones^[10]. The word that matters is structured. The same questions for every candidate, a written rubric, and interviewers who score against it the same way. That's where the predictive power comes from, and none of it depends on the interviewer being able to write the code.

Most engineering loops don't run that way. An engineer gets the invite, picks a question they like and judges the answer against what they would have done. Left to their own judgment, many end up answering whether they'd like working with this person. That's a fit question. The role needs to know whether the candidate can do the job.

The scores show it. One analysis followed engineers through several technical interviews, each with a different engineer interviewing them. Only about one in four got consistent scores, and candidates who were strong on average still failed a single interview up to 22% of the time^[11]. When the same person can pass or fail depending on who's asking, the loop is measuring the interviewer as much as the candidate. That noise also adds hours. A screen that passes the wrong people sends them onsite, and every extra onsite multiplies the hours above.

That says nothing bad about engineers. Most were never trained to interview, and nobody hired them for it. We made interviewing an unpriced second job for the most expensive people in the company, and then act surprised that it's inconsistent.

What an engineer catches

There are things an engineer in the room catches that nobody else does. They can push past a rehearsed answer into territory the candidate hasn't seen. They can tell a confident wrong answer from a right one. They can judge whether a design tradeoff holds up for the system you actually run. And strong candidates often decide to join because of one good conversation with the people they'd work with.

All of that argues for one engineer conversation, at the end, with a candidate who has already shown they can do the work. That's where an engineer's judgment changes the outcome. The candidate walks through something they built, and the engineer asks why each choice is there and what breaks when a requirement changes^[12]. It's also where the hardest part to hand off lives. When a candidate says something unexpected about an RF front end or a firmware race condition, the right next question usually isn't on any list.

Everything before that conversation is where the hours go: the phone screens, most of the onsite rounds and the debriefs that follow. Some companies pay outside engineers to run the first round. That saves their own engineers' time, but it still puts an engineer in a seat where the structure is doing most of the work.

Some people conclude that recruiters can't screen technical roles at all. One staffing founder told us he stopped using recruiters because they weren't technical enough. He's right about the gap. A recruiter on their own can't tell a strong answer about combustion instability in a rocket engine from one that only sounds right. But you don't need a rocket scientist to run the first interview for a rocket scientist. You need someone who knows which questions to ask and what a good answer sounds like.

The judgment you can hand off

What an engineer adds to a first-round screen is domain judgment. They know which questions matter for the role, what a strong answer covers, and which follow-up separates someone who has done the work from someone who read about it. Much more of that can be written down than most teams assume. Vercel's recruiting team took transcripts from its best engineer-interviewer, lined them up with how he had rated each candidate, and pulled out what his strongest yeses had in common. They framed the problem first, asked questions before solving, weighed short-term against long-term tradeoffs and thought out loud. It took a few hours^[13].

Once the judgment lives in the rubric, the person running the screen gathers evidence and the hiring team makes the call^[14]. The person gathering it doesn't have to be the expert.

That's why we built Pairwise. Before a technical screen, the recruiter gets an interview guide built for the role, with the questions to ask, what a strong answer should include and follow-ups for when an answer stays shallow. Afterward, they get an answer-by-answer evaluation of the transcript. The recruiter runs the conversation and decides who moves forward. Engineers spend their time on the final conversation with the few candidates worth it.

We hire this way ourselves. On our roles over the last year and a half, recruiters ran 80% of the interviews, and engineers met only the one in nine screened candidates that recruiters advanced. Of those, 46% passed the engineering round. When a recruiter rated a candidate a clear yes, the engineer agreed 83% of the time^[15].

Next time you're running a technical interview loop, ask what it's costing to have all those engineers in it. If recruiters ran every round but one, and an engineer met each finalist once, engineering time per hire would drop from about 70 hours to about 20. After paying for the recruiters' time, that saves about $5,000 per hire at a typical company and about $12,000 at big-tech pay^[16]. A team hiring 20 engineers a year keeps around $100,000 of engineering time.