Candidates
Design an interview people finish.
The choices that decide whether candidates complete an async interview, and how to find out where they give up.
7 min read · Updated September 24, 2026 · The LessRounds team
An asynchronous interview removes the worst part of an early-stage process for candidates: waiting for a slot. They can take it tonight, on their own schedule, without taking time off work. But it also removes the interviewer who would normally put them at ease, explain what’s happening and notice when they’re confused. Everything that person did has to be designed in. This guide covers the choices that matter most and how to measure whether they’re working.
Length: shorter than you think
Every question you add costs you some candidates, and the ones you lose are often the ones with the most options. Set the length from the job, not from how much you’d like to know.
| Role | Suggested length | Questions |
|---|---|---|
| Frontline and hourly roles | 10 to 15 minutes | 4 to 6 |
| Graduate and early-career roles | 15 to 20 minutes | 5 to 7 |
| Experienced professional roles | 20 to 30 minutes | 6 to 8 |
Each question should earn its place by testing a skill you’ve decided matters. If two questions test the same thing, cut one. Save “tell me about yourself” for a later round.
Say the length up front and keep to it. A candidate told 15 minutes who is still going at 25 will feel misled, even if the questions were good.
Device: meet candidates where they are
Many candidates, especially for hourly roles, will take your interview on a phone. Design for that.
- Choose audio-only when video adds nothing. For phone-based and customer-facing roles, what someone says is the job. Audio-only needs just a microphone, works on any phone, and is easier on patchy connections.
- Use video when seeing the person matters for the role, and tell candidates in advance that they’ll need a camera and somewhere they’re comfortable being seen.
- Check devices before the interview starts. A camera or microphone problem should show up at the start, not in the middle of question three.
- Don’t ask for an app download. Every extra install step loses people.
Clarity: remove every surprise
A nervous candidate with nobody to ask will fill any gap with worry. Before they start, tell them:
- Who they’ll be talking to, including that it’s AI if it is.
- How many questions and roughly how long.
- Whether they can pause, restart or re-record. If they can’t, say so clearly.
- That it’s recorded, and who reviews it.
- What to do if something goes wrong, with a real contact address.
Write questions the way you’d say them. One question at a time, one idea per question, no stacked sub-questions. “Tell me about a customer you helped who was upset. What did you do?” is clear. “Describe your approach to customer satisfaction, including examples, metrics and lessons learned” isn’t.
Feedback timelines: say when, then keep to it
Candidates will forgive a no. They rarely forgive silence. Put a date on your reply before they start (“We reply to everyone within a week”) and keep it, even if the answer is “we’re still reviewing.”
A simple standard that works for most teams:
- A thank-you screen straight after the interview that says what happens next.
- A reply to everyone within a stated number of working days.
- A clear, kind message to candidates who aren’t moving forward. The rejection email template is a starting point.
Accessibility: plan for it before anyone asks
Some candidates need adjustments, and an async interview should make that easy.
- Offer a way to ask for an adjustment, such as extra time or a different format, and tell candidates how in the invitation.
- Offer the interview in the language the job is done in. If the role is done in Hindi, screening in English tests the wrong thing.
- Keep instructions readable. Short sentences, plain words, and no instructions hidden in small print.
- Allow a reasonable window. A deadline of several days lets candidates choose a quiet moment. A 24-hour window punishes anyone with shift work or caring duties.
Measuring drop-off with a funnel
You can’t fix what you can’t see. Track every link as a funnel with five stages:
Opened → Started → Completed → Passed → Shortlisted
Each gap tells you something different.
| Gap | What it usually means | What to try |
|---|---|---|
| Opened but didn’t start | The intro screen or the ask put people off | Shorter intro, clearer length, mention the AI in the job ad so it isn’t a surprise |
| Started but didn’t finish | Too long, a confusing question, or technical trouble | Cut questions, reword the hardest one, offer audio-only |
| Completed but few passed | The bar or the questions don’t match who applies | Check the expected level and the ad’s targeting |
| Passed but few shortlisted | Reviewers disagree with the scoring | Revisit skills and weights with the hiring manager |
Here is a worked example with made-up numbers. Say 400 people open a link, 240 start and 150 finish. The open-to-start gap loses 160 people before they’ve answered a question, so the intro is the first thing to fix, not the questions. If you then shorten the intro and start rises to 300 out of 400, you’ve gained 60 candidates without spending anything on the ad.
Change one thing at a time and give it a few days, so you know what made the difference.
How LessRounds handles this
In LessRounds, candidates open one link, confirm their email with a code, and take the interview in the browser with no app or password. Interviews run from 10 to 60 minutes with 3 to 15 questions, in video or audio-only, in English or Hindi. Candidates are told it’s AI and recorded before they begin, they never see scores, and every link shows the opened-to-shortlisted funnel described above.