Behavioral Interview Question Bank

Write questions only for skills on the role's list, one skill per question. A question that probes three skills at once produces a rating nobody can use. The US Office of Personnel Management (OPM), the federal government's HR agency, says a structured interview typically assesses between four and six competencies¹.

Take the skills from the job's must-haves, ideally from a requirement register that gives each one a test. Then use an interview process template to give each skill a single interviewer. Keep each interview's list short: 45 minutes fits only three or four behavioral questions with their probes.

Ask what the candidate did, not what they would do or how they see themselves. OPM says each question should draw out the situation, the actions taken and their impact. Its guide suggests superlatives such as most, last, worst and least¹ to steer candidates to one specific incident.

Name the person or thing involved, too. "A site manager who signs the invoice" produces answers you can compare, while "a customer" lets everyone tell a different kind of story.

A situational question is not a generic hypothetical: it poses a realistic dilemma from the job and asks what the candidate would do. OPM notes that behavioral and situational formats have both proven to be effective², with behavioral questions showing higher validity in highly complex work such as professional and managerial jobs. Situational questions are the better choice when candidates may never have faced the situation.

A first answer is usually a summary. What separates a strong answer from an adequate one sits a layer down: the dead end, the detail left out on purpose, the change made afterward. Write three or four probes per question, in order, and stop when the answer runs out.

OPM's guide sorts its sample probes under situation, action and outcome, and asks interviewers to use very similar probes for all candidates¹. Improvised ones break the comparison: one candidate gets rescued by a good follow-up, and the next doesn't.

An anchor describes what an answer at each rating contains, so two interviewers give the same answer the same score. OPM's guide calls for example behaviors at every level, written to clearly differentiate between proficiency levels¹, and for at least three levels, ideally five to seven. This template uses four, so there is no middle score to settle on.

Draft the top and bottom first: the specific detail you'd be glad to hear, and the answer that explains the idea instead of describing an occasion. Then test the anchor on two real answers, one strong and one weak. If both land on the same number, it isn't finished.

In the US, the Equal Employment Opportunity Commission (EEOC) says pre-employment questions should be limited to what is essential for determining if a person is qualified³. It treats race, sex, national origin, age and religion as irrelevant to that judgment, and employers may not ask about disability before a job offer.

State law can go further. California, for example, generally bars employers from seeking an applicant's salary history⁴. Leave out brainteasers too, for a different reason: there's no real event in the answer to rate.

OPM's guide suggests a trial run with colleagues before a question goes live, to make sure it will draw an appropriate range of responses¹. Then keep checking in real interviews. Once a question has been asked 12 times, count its ratings. If one score takes 80% or more, the question isn't separating candidates, and this template calls it flat.

Exam writers check for the same thing. The University of Washington's assessment office notes that a test item has low discrimination⁵ when almost everyone gets it right, or almost everyone gets it wrong.

The second check needs at least eight hires with a performance review on the skill (this template's floor). Compare the reviews of those the question scored 3 or 4 with those it scored 1 or 2. If the low scorers did better, suspect the anchor, the interview's answer key. The same University of Washington page notes that an item strong students miss more often than weak ones may have been mis-keyed⁵.

Don't keep a question just because everyone asks it. The federal Uniform Guidelines on Employee Selection Procedures cover interviews⁶, and they rule out data bearing on the frequency of a procedure's usage⁷ as evidence of validity.

Retire a flat question and write its replacement in the same sitting, aimed at what failed. Mark the old one retired with the date and its numbers instead of deleting it, because someone will suggest it again.

After each round, a hiring debrief tool sorts the panel's comments by skill and shows what nobody tested. The template on this page holds the bank, the anchors and the counting sheets, and River can read your past scorecards to find the flat questions for you.