Twenty Minutes Buys Three Follow-Ups, Not Six Topics

How long should a phone screen actually run?

Twenty to twenty-five minutes for the round as most teams run it, and the number is what the turns add up to. Two minutes to frame the call, one opening question, three follow-ups on the answer, five minutes for the candidate's questions, and a minute to say what happens next. That is twelve or thirteen minutes of actual substance inside a twenty-minute slot.

The familiar fifteen-to-thirty band is unhelpful for one reason: it arrives attached to a topic list that cannot fit inside it. Background, skills, motivation, salary, availability and a company pitch is six subjects. Twenty minutes divided six ways is about ninety seconds each, which is exactly long enough to receive one prepared paragraph per subject before the clock moves you on.

That band was set when the call's job was to confirm a document that was itself the evidence. The resume did the work and the screen walked its headings. A candidate who has read any interview-preparation page now arrives with a tidy answer ready for every line of it, and the same drafting tools that tidy a resume will tidy an answer to any question anyone can see coming. Ninety seconds collects that answer and nothing else. The fourth minute on one answer is where something unrehearsed appears, and a six-topic agenda never reaches a fourth minute.

So the question to ask about a screening call is not how many minutes it has, but how many turns it can afford. Turns are the scarce unit, and everything on the agenda is competing for them.

What has to fit inside twenty minutes?

One claim, one frame, one close, and nothing else. The frame is two sentences: what this call decides and what happens after it. The close is five minutes of the candidate's questions, and it is the piece that gets cut first when the agenda overruns. Everything between those is one claim and the follow-ups on it, and there is room for exactly one.

Keep the opening pleasantries short and score nothing from them. In a study of 189 accounting students in structured mock interviews, the interviewer's impression formed during the rapport-building small talk before any structured question correlated .42 with that same interviewer's later structured score, falling to .25 and .24 when a different interviewer supplied the score ³. Those are mock interviews with students, and the authors trace the carry-over to rated competence, so the study is not evidence that minds are made up in the first three minutes. The measurement stands either way: the unstructured opening reaches the structured score even when a different interviewer writes it.

The frame is worth its two minutes because it buys back time later. Telling a candidate that the call is about one thing, that you will follow up on their answer rather than move on, and that pay and logistics are settled elsewhere, removes the pressure to volunteer a tour of the resume. Screens overrun when nobody says what the call is for.

And the candidate's five minutes are not a courtesy. They are the only part of the round where you learn what the person is actually trying to find out about the job, which is a different signal from anything they will tell you about themselves.

Cut topics before you cut turns

When the agenda does not fit, delete an agenda item and keep the exchange whole. A follow-up question is the only move in a screening call that a candidate cannot prepare from the posting, because it is built out of what they just said. Shortening turns to fit six subjects converts the whole round into a sequence of opening answers, which is the resume read aloud.

Structure is what buys prediction here, and it is not the same thing as coverage. In the 2022 re-analysis of the selection literature, structured interviews estimate at .42 against .19 for unstructured ones, where structured means every candidate is asked the same questions and rated on a common scale ². Those are corrected correlations with supervisor ratings pooled across many jobs, not accuracy rates, and the larger figures still in circulation belong to an earlier set of estimates that this one replaces, so the two never go in the same sentence. What a screening call can act on is the definition itself: the same questions, rated the same way.

A workable cut list, in the order things should go:

• Availability, notice period, work authorization, location. Form fields. They are faster to collect and easier to compare in writing.
• Compensation range. The posting, where a candidate can read it before agreeing to the call.
• The company pitch. Move it to the round after this one, when there is a reason to deliver it.
• "Walk me through your resume." It reproduces the document and eats four minutes.

What survives the cut is one claim, chosen before the call rather than during it, and the turns to pursue it. The mechanics of that pursuit are in what follow-up questions actually expose.

Why is adding ten minutes the wrong fix?

Because the binding constraint is recruiter capacity, and it has already tightened. Ashby's benchmark across more than a hundred million applications and 247,000 jobs reports the average recruiter processing 291 applications per hire against roughly 100 in early 2021, with the share of applications reaching an interview falling from around seven or eight percent to between 3.6% and 4.7% depending on role type ¹.

That is one vendor's customer base, skewed toward venture-backed technology employers, and application records include duplicate and automated submissions, which inflates the numerator by an amount nobody has measured. It is a direction rather than a level. The direction is enough: ten extra minutes on a week of thirty screens is five hours, and those hours come out of sourcing, out of debriefs, or out of the candidates nobody got back to.

The better trade is to spend the same twenty minutes on fewer things, and to move anything that can be answered asynchronously out of the call entirely. Skills evidence in particular does not need a longer conversation to exist; testing for AI skills without adding an hour to the loop works through the options.

If your team records screens with the candidate's consent, pull the last one and count the follow-up questions, meaning questions built out of something the candidate had just said rather than read off your list. If the count is zero or one, the round's problem was never its duration, and ten more minutes would have produced ten more minutes of the same thing.

Decide once what happens to the recovered minutes. A screen that reliably finishes at eighteen minutes is only a saving if the slot shrinks. Left at thirty, it fills.

See a sample report >