Asynchronous video interview best practices

Most advice on asynchronous video interviews is written by the companies selling the software, and it contradicts itself. One guide says keep it to three or four questions. The next says five to seven. One says a 24-hour deadline keeps things moving. The next says give a week. This page resolves those open questions with a defensible recommendation for each, and grounds the calls in what recruiters and candidates actually report.

The short answer

Run a first-round asynchronous interview at four to six questions, 60 to 120 seconds of recording time each, with a deadline of at least 48 hours and one retake per question allowed. Score every answer against a written rubric, keep a human making the advance decision, and commit to watching the answers before you ask anyone to record them. Those five settings carry most of the difference between a screen people complete and one they abandon.

How many questions: four to six

The contradiction worth resolving first. Vendor best-practice guides split between “3 to 4 maximum” and “5 to 7,” which is not much help when you are staring at a blank template. The honest recommendation sits in the overlap: four to six.

Fewer than four and you cannot separate a real answer from a lucky one, or read more than a single dimension of the role. More than six and you have built homework. Completion falls fast as a recorded interview gets longer, because for the candidate it is a heavy investment before they have spoken to a single human. One recruiter put the dynamic plainly: “You should only add friction points if you want candidates to self select out of the process.” Six is roughly where a careful screen tips into a filter on patience.

The discipline is not the count, it is the mapping. Name the two or three things this stage has to prove before you write anything. Communication, role-specific judgment, maybe one motivation check. Then write one question per thing, plus one or two to go deeper on the hardest one. If a question does not map to something you named, cut it. A tight five that each earn their place beats a padded seven every time. For prompts you can adapt, see questions to ask.

How long per answer: 60 to 120 seconds

Recruiters who configure these tools tend to land on 30 to 90 seconds of think time before recording starts, and 60 to 180 seconds of recording per answer. For a first-round screen, 60 to 120 seconds of recording time is the useful middle. It is long enough for a real example with a result, short enough that nobody is filling silence, and short enough that you can review the whole thing at speed later.

Give a short, explicit think window before the timer runs. A countdown that starts the instant a question appears is one of the mechanics candidates fear most, and the fear shows up in the recording, not in the answer.

How long a deadline: 48 hours minimum, 72 is better

This is the cheapest win in the whole format and the one most often gotten wrong. Give candidates at least 48 hours to complete the interview. Seventy-two is better. They have jobs, time zones, and lives, and the ones worth hiring are often the busiest.

A same-day or 24-hour deadline does not filter for ability. It filters for who happened to be free and checking email. That is friction pointed at the wrong target, and friction is exactly what drives the drop-off recruiters complain about: “Candidate drop off for a one way video interview is huge. For them it’s a fairly heavy investment before they’ve even spoken to someone.” A generous window costs you a day or two of calendar time and protects the completion rate you are about to depend on. One TA leader at a Fortune 100 reported a 50% take rate on their one-way interviews even at that scale, so every avoidable point of friction matters.

Retakes: allow at least one

Whether a candidate can re-record is a setting you control. As one recruiter explained, “The company can customize everything, like if a candidate can re-record their answers, or if a candidate can review their submitted responses.” Use that setting. Allow at least one retake per question.

A single-take, no-retake format is the single most common thing candidates report panicking over. One described freezing on the first question of a sales screen: “I didn’t notice the time limit, and there were no retake options,” after applying to more than 300 entry-level roles. A first-question stumble, a barking dog, a misread prompt, none of that tells you anything about whether the person can do the job. One retake removes the worst of that anxiety. It does not let anyone script a flawless performance, because re-recording every answer to perfection takes more time and effort than almost any candidate will spend, and the polish is obvious when they do. If you want to limit it, cap retakes at one or two per question rather than banning them. For the candidate-facing version of this, see how many retakes.

Write questions that cannot be scripted

The format only works if the questions force a real answer. Anything a candidate can recite from a “top interview answers” article tells you nothing, because everyone in your pool read the same article.

• Weak: “Are you a good communicator?”
• Strong: “Tell me about a time you had to explain something technical to someone without the background. What did you do, and how did you know it landed?”

Ask for a specific example, a decision and the reasoning behind it, or a short response to a realistic scenario from the actual role. Scenario prompts work especially well, because they show how someone thinks rather than what they have memorized. Keep the wording plain and the same for everyone. Standardizing the questions is what lets you compare answers fairly later, which is most of the case for using this format at all.

Score against a written rubric, not a gut feeling

Decide what a good answer looks like before you watch a single one. A simple rubric, one to four points on each of two or three traits, keeps reviewers honest and makes answers comparable across a large pool. It also gives you a defensible record of why one candidate advanced and another did not, which matters more every year.

Without a rubric you rebuild the bias the structured format was supposed to remove. With one, a recorded screen can be more consistent than a phone call, because everyone answered the same questions under the same conditions and got rated on the same scale. Have a second person score the borderline cases when you can. Disagreement is signal, not noise. The mechanics of building and applying a scale live in how to score async interviews.

Actually watch the answers

This is the part vendors skip and the part that quietly decides whether the whole exercise was worth anyone’s time. If you ask people to record answers, watch them.

The most damning recruiter feedback is not about the tools at all. It is about review discipline. One described it bluntly: the format “doesn’t allow for probing and I found that most of the video responses weren’t even seen, hiring managers ignored them and just based decisions on who to move forward off the resume.” If that is what happens, you have spent your candidates’ time, taught them to resent your brand, and learned nothing you did not already have on the resume. The same recruiter who valued the format, the one who said it surfaced strong communicators with weak resumes she would otherwise have missed, only got that value because she watched.

Make review fast so it actually happens. Watch at 1.5x, read the transcript, score against the rubric, and move the shortlist into live conversations within a day or two. Tools that transcribe, summarize, and surface the parts that match your criteria help with the triage, and that is a fair use of automation. The advance decision should still belong to a person who read the answers against the rubric. A fast, respectful screen is a real advantage. A pile of recorded interviews nobody opens is worse than the phone call you replaced.

Protect the experience, and say why up front

Two or three small things separate a screen that feels respectful from one that feels like a wall:

• Tell candidates why you use this format and how long it will take, in the invitation, before they start.
• Offer a practice question so the real one is not the first time they meet the tool.
• Let people know when to expect to hear back, and then do, including the no.
• Offer an alternative format for anyone who needs an accommodation, and make that offer visible rather than buried.

Recruiters who run high volumes are split on the format for a reason, and the split tracks how it is run. The teams whose candidates “NOPE the hell out” tend to be the ones using long, no-retake, tight-deadline screens that nobody watches. The teams who get value tend to keep it short, give people room, and read every answer. The settings on this page are the difference.

If your goal is a screen people finish rather than rage-quit, the companion piece is how to run a one-way interview candidates don’t hate. When you are ready to pick a tool to run all of this, the software comparison walks through the main options.