You get a link, a deadline, and a page with a red record button. No interviewer, no small talk, no nodding face to tell you whether that last answer landed. One-way video interviews (HireVue, Spark Hire, Willo, VidCruiter and friends) are now the default first filter at a lot of large employers, and most candidates walk into them assuming the advice from live Zoom rounds carries over. It mostly does not.
Live calls give you feedback loops: you course-correct, ask a clarifying question, read a raised eyebrow. Asynchronous rounds strip all of that out and replace it with a stopwatch. That changes what a good answer looks like.
How these platforms actually work
Before you can prepare, you need to know what the machine will do to you. The mechanics vary by vendor but the shape is consistent:
- A fixed prep timer. You see the question, then get 30 seconds to 2 minutes to think before recording starts automatically. Some setups skip this entirely.
- A fixed answer window. Usually 60 to 180 seconds. When it ends, the recording cuts, mid-sentence if that is where you are.
- Limited retakes. Some platforms give you two or three attempts per question. Some give you one. Some let you re-record the whole set once. The invite email or the practice screen tells you which, and it is worth reading properly.
- A one-way street. Once you submit a question, you cannot go back and improve it.
- A total deadline. Typically 3 to 7 days from the invite, often with automatic rejection if you miss it.
Read the invite for two numbers: seconds per answer and retakes allowed. Those determine your whole strategy. Zero retakes means you rehearse before you open the link.
The pre-flight checklist
Technical failure is the most common self-inflicted loss in this format, and it is entirely preventable. Run this the day before, not five minutes before.
- Test your upload speed, not just download. These platforms push video up. Anything under 3 Mbps up is risky, and hotel or cafe Wi-Fi will fail you.
- Use the browser they recommend, usually Chrome or Edge on desktop. Safari and in-app browsers cause the most support tickets.
- Grant camera and microphone permissions in advance and confirm the preview shows both working. Check your OS-level privacy settings too, not just the browser prompt.
- Close everything else. Slack, Spotify, Dropbox sync, backup software, twelve tabs. They compete for bandwidth and CPU, and a notification sound will land in your recording.
- Plug in. Laptop on mains, phone charged, and a hotspot ready as a fallback.
- Have a second device signed in to the same invite link if the platform allows it. If the laptop dies mid-round, you switch rather than email support and hope.
- Do the practice question if there is one. It is free, it does not count, and it shows you the exact timer behaviour.
Structuring an answer for 90 to 120 seconds
Two minutes is shorter than you think. Most people spend the first 30 seconds warming up, then hit the time limit halfway through the interesting part. Front-load instead.
Open with your answer, not your context. "I led the migration of our billing system, and the hard part was the data reconciliation" beats "So, at my previous company, which was a fintech of about 200 people..." Then give one concrete situation, what you specifically did, and the outcome with a number if you have one. Close with a single sentence, and stop talking. Silence at the end is fine. Rambling to fill time is not.
The unit of preparation is bullet points, not sentences. Write three or four keywords per likely question on a sticky note placed just below the webcam. A memorised script reads as a memorised script, and if the timer cuts you off in the middle of it, you have nothing to fall back on. Bullets let you land the same points in whatever time you actually have.
Talking to a lens with nobody nodding back
Look at the camera, not at your own face on screen. Your eyes will drift to the preview window because faces are magnetic; move the browser window so it sits directly under the lens, or hide the self-view if the platform lets you.
Smile at the start of each answer, deliberately. Not a grin, just the softening you do when you greet someone. Without a listener your voice flattens fast, so push slightly more energy than feels natural, and speak marginally slower than you would on a call because compressed video eats consonants. Use your hands in frame; stillness reads as tension.
If watching your own recording feels physically unpleasant, that is normal and it is not a signal about the quality of the answer. Judge it on whether the point was clear in the first fifteen seconds.
When the timer beats you and other failures
The prep timer starts before you are ready, or you sneeze, or the recording freezes. Have a plan.
If the timer starts and your head is empty, say one framing sentence out loud while you think: "There are two examples I could use here, and the more useful one is the pricing project." That buys eight seconds and sounds deliberate. If you are cut off mid-sentence and retakes are gone, do not panic about it, most reviewers watch on 1.5x and are scanning for substance rather than polish.
If the recording genuinely fails (upload error, black video, no audio), email the recruiter and platform support immediately with a screenshot and a timestamp. Same-day contact almost always gets a reset. Silence for two days does not.
The questions that come up almost every time
These rounds are short, usually four to six questions, and they are near-standardised:
- Tell me about yourself, or why this role. Ninety seconds, not your whole CV.
- Why this company specifically. This is where generic answers die, so name something real about the product or the market.
- A behavioural one: a conflict, a failure, a time you missed a deadline.
- A relevant strength or a technical basic tied to the job description.
Prepare those four with bullets and you have covered most of what any vendor will throw at you. If you are pulling those examples from a CV you tailored per application, keeping a single structured record of your own achievements helps, which is essentially what Postulit builds out of your LinkedIn profile.
About the AI scoring, honestly
Some vendors run automated analysis on speech patterns and word choice. Others just record and route to a human. HireVue publicly dropped facial analysis in 2021, and regulation in Illinois, New York City and the EU has pushed the industry towards disclosure and away from opaque visual scoring. You usually cannot tell from the outside which pipeline you are in.
So do not try to game a model. Clarity, steady pacing and relevant language from the job description serve you either way. Those are the same things that make an answer good anyway.
The practice protocol
Fifteen minutes, the night before. Open your phone camera, ask yourself "tell me about yourself", record for 90 seconds, stop. Watch it once. Pick exactly one thing to fix, usually pace or a filler word or where you were looking, then do one more take.
That is it. Do not do eight takes and do not transcribe yourself. The single throwaway practice run is what separates candidates who look composed from equally qualified ones who treat the round like a form to submit as fast as possible. Two takes on your phone, tonight, and you are ahead of most of the pile.