A hiring manager I used to work with had a favorite interview question: "If you were an animal, which animal would you be?" I sat in on one of his interviews once, and in the debrief afterwards we got to the candidate's answer. She had said a bear. I remember thinking, as a Finn, what a magnificent answer that was: in our folklore the bear is the god of the forest, calm around people, furious only in defense of her cubs, strong, fast, honestly a little cute. Then my boss gave his verdict. "Bears are slow and they hibernate. I bet she's lazy."
One answer, two interviewers, two confident and opposite readings, and not a word of it about the job. A question with no connection to the work doesn't measure the candidate at all; it hands the interviewer a mirror. The fix starts with knowing what each type of question can actually measure. This article covers the three types of interview questions worth asking candidates, what each one is for, and the evidence behind the default this whole site is built on: ask about past behavior first.
The three types of interview questions
Once your job analysis has named the competencies to test, every question you could ask a candidate falls into one of three working families:
Behavioral questions ask for a specific episode from the candidate's past. "Tell me about a time you had to deliver on a deadline you knew was unrealistic." The premise is the most durable finding in selection research: the best predictor of what someone will do is what they have done. The candidate supplies a real event, and you score how they actually behaved.
Situational questions pose a hypothetical and ask what the candidate would do. "A key customer threatens to leave over a mistake that wasn't your fault. What do you do?" The premise here is that intentions predict behavior: how people say they would act reveals what they understand the situation to demand.
Knowledge questions check whether the candidate knows things the job assumes. "What does GDPR require before you can email a list of prospects?" No story, no hypothetical, just a fact or a procedure the role cannot function without.
All three are legitimate structured-interview material; the classic taxonomy of interview structure (Campion, Palmer and Campion, 1997) names exactly these as the "better question types" that structured content is built from. Notice what the list quietly excludes: brainteasers, animal questions, strengths-and-weaknesses rituals, and culture-fit vibes questions. None of those has a premise that connects the answer to the job, which is another way of saying they predict nothing.
The three families are not interchangeable, and they are not equal. The evidence has a clear favorite.
Why behavioral questions are the spine
When researchers put the two main families head to head, past behavior won. Taylor and Small's meta-analysis (2002) compared structured interviews built on behavioral questions against those built on situational ones: behavioral questions predicted job performance at r = .56 against .45 for situational. Both numbers are good. One is better, and the gap is not noise.
The more interesting finding is where the gap widens. Situational questions hold up for simpler, well-defined work and weaken as job complexity rises. That makes intuitive sense: anyone can narrate an ideal response to a hypothetical, and the more senior the role, the better-rehearsed the ideal. A real episode is harder to fake. It comes with details a hypothetical never demands, what the candidate split, what they parked, who pushed back, what it cost, and each of those details is something a planned follow-up probe can test.
Here is what a behavioral question looks like when it arrives as a full recipe card rather than a bare question:
Everything on that card exists because the question type alone is not enough. Taylor and Small's other finding is easy to skip past: both question types improved markedly when answers were scored against descriptively anchored criteria. A behavioral question scored on gut feel gives back much of its edge. The question is the ingredient; the anchored scoring is what turns it into a meal.
When situational questions earn their place
If behavioral questions win, why keep situational ones at all? Because the behavioral premise has a precondition: the candidate needs past behavior to describe. A new graduate applying for their first support role has never owned an escalation. Asking them to "tell me about a time" produces either silence or a story stretched thin from a summer job, and neither tells you much.
Situational questions were invented for exactly this. Latham and colleagues built the situational interview (1980) by taking critical incidents from job analysis, real moments where the work went notably right or wrong, and turning them into hypotheticals. A good situational question is not a creative thought experiment; it is a dilemma the job will genuinely serve, with the names removed. That connection to real incidents is what separates a situational question from a hypothetical built for drama, and from anything involving animals.
This is why every situational card in the pantry carries a seniority flag. Here is one, written for juniors, and note that it tests the same competency as the behavioral card above:
Same competency, two doors into it. The behavioral door needs a past; the situational door only needs judgment. Use situational questions for entry-level candidates and career changers, and for the occasional dilemma so central to the role that you want every candidate reasoning through it on equal footing. For experienced hires, let past behavior carry the interview.
What knowledge questions are for
Knowledge questions are the narrowest family and the easiest to overuse. They verify the floor, not the ceiling: does the candidate know the regulation, the method, the tool, the terminology the role assumes on day one? Where the job has hard prerequisites, a few direct knowledge checks beat hoping the resume is honest.
Two cautions. First, an interview minute spent on a fact is a minute not spent on behavior, and facts are cheap to verify elsewhere. A short work sample or practical task often checks knowledge better than conversation, which is why our job-analysis step recommends routing technical "can they do it" requirements to a practical check and saving the interview for what only an interview can reach. Second, knowledge questions age. The candidate who can recite your stack today and the one who has repeatedly learned new stacks are different hires, and only behavioral questions find the second one.
Interview formats: a shorter conversation than you'd expect
Question types are a bigger decision than interview formats, and the evidence keeps it that way. Whether you run one-on-ones, a panel, or a sequence of rounds, and whether it happens in a room or on a video call, matters far less than whether every candidate gets the same questions, scored against the same anchors, by interviewers who rate independently.
A few format choices do earn attention. More interviewers help, up to a point: Google's structured interviewing program landed on a "Rule of Four", finding that after four structured interviews, additional interviewers barely change the decision, which is an argument for a short sequence of focused one-on-ones over a long gauntlet. Panels can work too, if every panelist scores independently before anyone speaks; a panel that deliberates its way to one shared number is just a committee with better seating. And sequential rounds have a quiet advantage: they let you split competencies across interviewers, so each round goes deep on two or three instead of skimming six. That division of labor is the interview guide's job to fix in writing.
Format, in other words, is the table setting. The meal is the questions and the scoring.
Composing the interview
Put the families together and the recipe for question selection almost writes itself:
- Take the four to six competencies from your job analysis, and give each two or three behavioral questions. This is the spine, for every role, at every level.
- For junior candidates, swap in situational variants where there is no past to ask about. The pantry's seniority flags mark them.
- Route hard prerequisites to a work sample or a short knowledge check, and spend the interview minutes you saved going deeper on behavior.
- Score every answer against anchored criteria, whatever the question type. The type picks what you ask; the anchors decide what the answer is worth.
Every question in the pantry comes typed and flagged this way, with the probes and anchors already attached, so composing an interview is closer to picking ingredients than writing a cookbook. If you want the whole path in one guided pass, the planner walks you from job analysis to a printable plan in about three minutes, free, no account needed.
And the bear? Still the best answer I ever heard to a question that should never have been asked. If what my boss actually wanted to know was whether she would push through slow, unglamorous work, there was a question type for that: "Tell me about a time you had to keep something moving long after it stopped being interesting." She would have had a real answer. We never asked for one.