Ask questions only the real thing could answer.
“Tell me about your experience with X” gets a rehearsed answer from everyone. Questions anchored in a candidate's own CV claims separate the people who did the work from the people who watched.
Why generic questions fail
Every question bank on the internet has been read by your candidates too. Strengths and weaknesses, conflict stories, five-year plans — polished answers to all of them can be prepared in an evening, and they tell you who prepared, not who can do the job.
The alternative is claim-based questioning: treat each concrete claim on the CV as a hypothesis and design questions that are easy to answer if the claim is true and awkward to answer if it isn't. Three properties make a question work:
- Specific to their claim — it references the project, number, or tool they put on the page, so there is no stock answer.
- Asks for mechanism, not outcome — "what did you change and how did you measure it" instead of "was it successful".
- Probes ownership — "what part did someone else own?" is the quickest way to find out what they actually did.
Example question sets
Each set starts from a typical CV claim and shows the follow-ups that pressure-test it.
Software developer
CV claim: “Led migration from monolith to microservices, cutting deploy time 4×.”
- Which service did you split out first, and why that one?
- What was your rollback plan when a migrated service failed in production?
- Deploy time dropped 4× — from what to what, and how was it measured?
- What part of the migration did someone else own?
Marketer
CV claim: “Grew organic traffic from 5k to 60k monthly sessions in a year.”
- What were the top three pages driving that growth, and why did they work?
- What did you try that failed before you found what worked?
- How much of that traffic converted, and to what?
- What tooling did you use for keyword research and tracking — and what would you change?
Sales rep
CV claim: “Consistently exceeded quota, closing $1.2M in ARR last year.”
- Walk me through your largest deal last year — how did it start, and what nearly killed it?
- What was your quota, and how was it set?
- How much of your pipeline was self-sourced versus handed to you?
- Which deal did you lose that you should have won, and what did you change afterwards?
Doing this for every applicant is the hard part
Claim-based questions must be written per candidate — that's the point — and with 200 applicants that's a week of work. BestHire generates them automatically: every fit applicant takes a 5-minute AI interview with questions built in real time from your JD and their specific CV, and every answer is scored against the claim it was probing — with the transcript quote cited. See how it works.