Screening questions: what to ask a machine to verify

The screen is where roles are actually lost

Not at sourcing, and not at the final interview. Roles are lost in the gap where a candidate looked right, took up three people’s time, and turned out to have a constraint that was knowable in the first two minutes.

Screening exists to close that gap, and it is the stage most often handed to a machine. Which makes the question of what a machine can genuinely verify worth answering properly, rather than handing over the same list you would give a junior recruiter.

Three kinds of question, and only two belong here

A screening question is worth asking if a wrong answer would stop the process. Everything else is curiosity.

Write them so the answer is unambiguous

• Ask for the number, not the comfort level.“Are you flexible on salary?” produces a yes from everyone. “What figure would make you accept?” produces a number you can act on.
• Ask about the constraint, not the preference.“Would you consider relocating?” and “Can you be in Pune from January?” are different questions and only one of them screens.
• Separate compound questions. Notice period and buyout willingness are two facts. Asked together, you reliably get one answer.
• Avoid anything that invites a performance.“Why do you want this role?” at screening stage produces a rehearsed paragraph from everyone and separates nobody.

Four is usually the right number

Long screening lists degrade in two ways at once. Candidates start answering to get through it, so the later answers are worse than the early ones, and a screen that feels like a form is the first place a strong candidate disengages, because they have other options and this one is already tedious.

Pick the constraints that would genuinely stop the process for this role. On most roles that is location, notice, money, and one thing specific to the job.

The gap questions matter more than the standard ones

The standard set is the same for everyone, so it tells you nothing about an individual beyond the constraint. The higher-value question is the one that only applies to this candidate: the requirement their profile never addressed, the two-year gap, the title that could mean two different jobs.

Doing this manually means reading every profile carefully enough to spot the gap and remembering to raise it. That is precisely the discipline that collapses at volume, and it is the part worth automating.

What to do with the answers

1. Treat a factual answer as decisive. If someone cannot be in the location and the role is genuinely onsite, that is the end, and continuing wastes both parties’ time.
2. Treat a probed answer as evidence for the human interview rather than a verdict. It tells you what to push on, not whether to hire.
3. Record everything against the candidate. The most common failure is a good answer given in a conversation nobody wrote down, rediscovered at offer stage.