Quick verdict: The best questions to ask an interviewee are the same five to eight job related questions asked of every candidate, each scored on a written rubric. The 2022 Sackett meta analysis puts structured interviews at a validity of .42, the highest of any common hiring method, against .19 for a free flowing chat. The question matters less than the structure around it.
What is on this page: 16 core questions sorted by competency, each with a follow up probe, what a strong answer sounds like for each, the red flags to note, a four level scoring rubric you can paste into your ATS, and the questions you should never ask.
Updated October 9, 2026. Research figures checked against the sources linked below.
What are the best questions to ask an interviewee?
The best questions are tied to a competency the job actually needs, asked the same way to every candidate, and backed by follow up probes you wrote in advance. A clever question asked once tells you very little. A plain question asked of six people and scored against the same anchors tells you who is strongest.
Google describes its own approach in four parts: vetted questions relevant to the role, detailed notes on each answer, scoring on standard rubrics that define poor, borderline, solid and outstanding, and interviewer training and calibration. That is the model this page follows, and it is laid out in the Google re:Work guide.
Start by picking the four to six competencies that matter most for the role. SHRM organises its own question library the same way, telling interviewers to choose the most important attributes first and then gather questions for each, as its interview questions toolkit explains. If you have not agreed those competencies with the hiring manager yet, run an intake meeting first.
Do structured interviews really predict job performance better?
Yes, and the gap is larger than most interviewers assume. In 2022, Paul Sackett and colleagues reworked the statistical corrections used in decades of hiring research and found that structured interviews had the highest mean operational validity of any predictor, at .42, according to HumRRO’s summary. Cognitive ability, long treated as the top predictor, came in at .31.
The older figures still circulate. Schmidt and Hunter’s 1998 synthesis gave .51 for structured interviews and .38 for unstructured ones; Sackett’s revision gives .42 and .19, which makes structured interviews about twice as predictive, as Test Partnership notes. The chart compares both sets.
One caution keeps this honest. The .42 is an average, and studies vary widely around it. The 80% credibility interval for structured interviews runs from .18 to .66, according to Master International. In plain terms, a well designed structured interview can be excellent and a sloppy one can be no better than a chat. The design work below is what moves you toward the top of that range.
Which hiring methods predict performance best in 2026?
If you are deciding where to spend interview time, the revised table is a useful map. Methods built around the specific job rank highest, and general signals like years of experience rank lowest.
Notice where years of experience lands. KiTalent’s reading of the revised table puts it at .07, practically nothing once a candidate clears the basic competence bar, as its research chapter sets out. A question like “How many years have you done this?” is the weakest thing you can ask.
What should I ask to test problem solving?
Problem solving questions work best as hypotheticals with a realistic twist, followed by probes that push on the reasoning rather than the answer. You are scoring how the candidate breaks the problem down, not whether they reach your preferred solution.
| Question | Probe to follow with | A strong answer sounds like | Red flag |
|---|---|---|---|
| Our biggest customer says our last delivery was late and wrong. You have one hour before their call. What do you do? | What do you need to know first? Who else do you bring in? | Separates fact finding from fixing, names the data they would pull, plans what to say if the cause is still unknown | Jumps straight to apologising or blaming another team |
| Tell me about a problem you solved where the obvious fix was wrong. | How did you notice? What did it cost to change course? | Explains the signal that showed the first fix failed and what they tested instead | Cannot name a time they were wrong |
| Walk me through a decision you made with incomplete data. | What would have changed your mind? | States the assumption, the risk accepted, and how they checked it later | Claims they always wait for full information |
| If you had to cut this team’s budget by 15% next quarter, where would you look first? | What would you protect, and why? | Asks what the goals are before cutting, ranks options by impact | Cuts evenly across everything |
What should I ask to test communication and collaboration?
Collaboration answers are easy to rehearse, so the probes do the work. Ask for one specific person and one specific disagreement, then ask what the other person would say about it.
| Question | Probe to follow with | A strong answer sounds like | Red flag |
|---|---|---|---|
| Tell me about a time you disagreed with a colleague on how to do the work. | What did they want? What did you concede? | Describes the other view fairly and says what changed as a result | The colleague is a villain in the story |
| Describe a time you had to explain something technical to someone outside your field. | How did you know they understood? | Mentions checking understanding, not just talking | Says people “just did not get it” |
| Give an example of feedback you received that was hard to hear. | What did you do differently afterwards? | Names the feedback plainly and a concrete change | Only “I work too hard” style answers |
| Tell me about a project where you depended on another team that was slow. | What did you do before escalating? | Builds the relationship first, escalates with evidence | Escalated on day one, or never |
What should I ask to test ownership and drive?
| Question | Probe to follow with | A strong answer sounds like | Red flag |
|---|---|---|---|
| Tell me about something you improved that nobody asked you to fix. | How did you get others to adopt it? | A clear before and after, with a measure | Every example was assigned to them |
| Describe a goal you missed. What happened? | What would you do differently now? | Owns their part without excess self blame | Only external causes |
| Tell me about a time you had too much on and had to drop something. | Who did you tell, and when? | Prioritised openly and warned people early | Quietly let things slip |
| What is the most useful thing you learned in the last year, and how did you learn it? | Where have you used it since? | Specific skill, specific use | Vague answer about “staying current” |
What should I ask to test role knowledge?
Google calls this role related knowledge and recommends behavioral questions here, to check how a candidate applied their skills in real situations rather than simply listing them, according to its re:Work guidance. Adapt these four to the role.
| Question | Probe to follow with | A strong answer sounds like | Red flag |
|---|---|---|---|
| Walk me through the last piece of work you would put in a portfolio. | What was your part, exactly? What would you redo? | Separates their own work from the team’s | Says “we” for every step |
| Which tool or method in your field is overrated, and why? | What would you use instead? | A reasoned opinion with trade offs | No view at all |
| How would you measure whether you were doing this job well after 90 days? | Which number would you watch weekly? | Names two or three metrics that match the role | Only effort measures like hours worked |
| Tell me about a mistake specific to this kind of work that you have seen others make. | How do you avoid it? | Shows depth through a concrete trap | A generic answer any role could give |
What questions should I never ask an interviewee?
Anything that reveals a protected characteristic. The EEOC says pre employment questions should be limited to what is essential to judge whether someone is qualified, and employers are explicitly prohibited from asking about disability before an offer, according to its guidance on inquiries. Avoid questions about age, family plans, religion, national origin, or club memberships that would reveal them.
| Avoid asking | Ask instead |
|---|---|
| Do you have children, or plan to? | This role needs travel two weeks a month. Is that workable for you? |
| Where are you originally from? | Are you legally able to work in this country? |
| How old are you? When did you graduate? | Walk me through the experience most relevant to this role. |
| Do you have any health conditions? | Can you perform the essential duties of this role, with or without reasonable accommodation? |
| What church or clubs do you belong to? | Which professional groups or communities help you in your work? |
How do you score interview answers fairly?
Write the rubric before the first interview, score each answer on your own before the panel talks, and record the evidence next to the score. A four level scale is enough, and it matches the four levels Google uses.
| Score | Label | What it means | Example evidence you would note |
|---|---|---|---|
| 4 | Outstanding | Specific, owned example with a measured result and clear reasoning; handled every probe | “Cut invoice errors from about 40 a month to 6 by adding a check before sending” |
| 3 | Solid | Specific example with a sensible approach; one probe answered thinly | Clear story, but result described as “it went well” |
| 2 | Borderline | General answer or a team story where their own part is unclear | Mostly “we”, no decision they made themselves |
| 1 | Poor | No relevant example, or an answer that contradicts what the job needs | Blamed others, or could not name an instance |
Panels should score independently, then discuss only the large gaps. The US Office of Personnel Management describes exactly that process for structured interviews, and notes that higher structure brings higher validity, rater agreement and less adverse impact, in its assessment guidance.
Copy and paste: interviewer scoring note
Candidate: [name] | Role: [title] | Interviewer: [name] | Date: [date]
Competency: [for example, problem solving]
Question asked: [exact wording]
Evidence heard (quotes or facts, no adjectives): [notes]
Score (1 to 4): [score] | Why this score and not one higher: [one line]
Red flags noted: [none, or what]
Submitted before the debrief: yes / no
Should I use behavioral or situational questions?
Use both, for different jobs. Behavioral questions (“Tell me about a time”) test what someone has done; situational questions (“What would you do if”) test how they reason through something new. OPM reports that behavioral description interviews have shown higher validity for highly complex work such as professional and managerial roles, according to its structured interview page. For entry level hiring, where candidates have less history, situational questions give more people a fair chance to show their thinking.
For a much longer behavioral list sorted by theme, see our behavioral interview questions guide.
How many questions should one interview include?
In a 45 minute slot, plan for five or six main questions. Each one needs time for the answer plus two or three probes, and you need ten minutes for the candidate’s own questions. Spread competencies across the panel so no two interviewers ask about the same thing; repetition wastes the candidate’s time and gives you the same evidence twice.
| Interview length | Main questions | Probes per question | Candidate questions |
|---|---|---|---|
| 30 minutes | 3 to 4 | 2 | 5 minutes |
| 45 minutes | 5 to 6 | 2 to 3 | 10 minutes |
| 60 minutes | 6 to 8 | 2 to 3 | 10 to 15 minutes |
These are Work Insiders planning ranges, not research figures.
How we built this page
Validity figures come from the 2022 Sackett, Zhang, Berry and Lievens meta analysis as summarised by HumRRO, Master International, KiTalent Research and Test Partnership, and the 1998 Schmidt and Hunter figures from the same secondary tables. Google’s process and its internal results come from its re:Work guide, updated March 2026. Legal guidance is from the EEOC. The question bank, probes, red flags, rubric anchors and timing ranges are Work Insiders editorial guidance built on that research, not figures from it. All sources were checked on October 9, 2026.
Quick answers
What is the single best interview question?
There is not one. The best question is whichever one targets a competency the job needs and is scored on a rubric. Asked that way, “Tell me about a time you disagreed with a colleague” beats any clever brainteaser.
How predictive are structured interviews?
Sackett and colleagues (2022) put their mean validity at .42, the highest of the common hiring methods they reviewed, against .19 for unstructured interviews.
Should every candidate get exactly the same questions?
Yes for the main questions, so answers can be compared. Planned probes can go deeper where an answer is thin, but write them in advance.
Is a five point rating scale better than four?
Not usually. Four levels, poor, borderline, solid and outstanding, matches Google’s rubric and stops raters parking everyone at three.
Do structured interviews feel robotic to candidates?
Google found the reverse: rejected candidates who had a structured interview were 35% happier than those who did not.
Can I ask about salary history?
Ask about pay expectations instead, and check your local rules first, because some jurisdictions restrict questions about past pay.
What is a red flag answer?
An answer with no specific example, one where the candidate cannot separate their own work from the team’s, or one where every problem was someone else’s fault.
Where can I learn the terms used here?
Our recruiting glossary defines structured interview, adverse impact and scorecard in plain English.
The bottom line
The questions on this page will work, but only inside a structure: the same questions for every candidate, written probes, a four level rubric, and scores recorded before the debrief. That structure is what lifts an interview from a .19 predictor to a .42 one.
A bad hire is expensive, so the hour you spend writing a rubric is cheap by comparison. Our breakdown of the cost of a bad hire shows what is at stake, and our skills based hiring playbook shows how interviews fit with work samples and tests.
Sources: HumRRO on Sackett 2022, Master International, KiTalent Research, Test Partnership, Google re:Work, OPM structured interviews, SHRM interview questions, EEOC prohibited practices
Free newsletter
Get the Work Insiders weekly
One email a week: the recruiting and HR tech news that mattered, the tools we just added to the directory, and one practical guide. No spam, and you can unsubscribe in one click.