Every async interview tool asks you how many attempts a candidate gets per question, and most teams pick a number once and never revisit it. It is worth about five minutes of thought, because the answer changes what you are measuring.
What each setting actually tests
One attempt measures composure under pressure and unrehearsed thinking. It is closest to a live conversation. It is also the setting most likely to eliminate a good candidate whose dog barked forty seconds in.
Two or three attempts measures the quality of a considered answer. Candidates settle, they fix the take where they lost their thread, and what you watch is closer to their best work than their most nervous minute.
Unlimited attempts measures preparation and persistence. Occasionally that is exactly the trait you want. Usually it just means you are watching a rehearsed performance, and the tenth take of everyone's answer starts to sound the same.
Choose per question, not per interview
This is the part worth using. A round can hold both:
- Warm-up: three attempts. Nobody should be filtered out on the first thing they said on camera.
- Experience question: two attempts. Room to restart, not room to script.
- The live-thinking question — how would you debug this, what would you check first: one attempt, and say clearly in the question why.
Tell them the number up front
A candidate who does not know how many takes they have will treat every take as their last, which reintroduces exactly the pressure you were trying to remove. Say it in the question text.
A note on fairness
Retakes are not an accommodation, they are a default. People interview from kitchens, shared flats and lunch breaks. A generous attempt count costs you almost nothing and quietly removes a class of noise that has nothing to do with whether someone can do the job.
