Reverse Interview Questions

A while ago I wrote about my values at work, and that if I had a choice, what would I choose. It is not often that I have that choice, but it is good to aim idealistic and expect realistic. So, in that line, I have also prepared a list of questions that I’d be asking in reverse interviews to get a signal regarding those values that I can expect at a job I’m offered but has not accepted yet. If you don’t know what a reverse interview is read about it first, I don’t have a good source handy.

There are two things worth noting first:

I’m going to ask the majority of these questions after I’ve got an offer to figure out the fit.
My expectation (based on my experiences) is that most of the answers to these questions are unsatisfactory, but at least will give me an idea of what I’m walking into.

Each row below is one question, who to ask it of, the signal I read in the answer, and what a strong or weak answer looks like. It is grouped by the person I am asking, then by what I am probing for.

ID	Question	Signal	A good answer	A weak answer
Teammate / peer IC (primary ground truth, the manager will be optimistic)
The work itself
T1	Tell me what you worked on over the last five working days, and what is on your plate for the next five.	High v3	a concrete account of what they did over five days and what is next. I do the classifying, so I read the grunge density myself.	"the usual" or nothing specific means either the work is unremarkable or they are guarding it. POLICY "lots of variety" carries no data.
T2	What is the single biggest bottleneck in your day to day that stops you doing your best data-science work?	High v3 v7	names a concrete friction (tooling, data quality, unclear priorities, compute, meetings) and says whether anyone has tried to fix it.	"nothing really, it is great" means it is not safe to say or they have low self-awareness. Blaming people by name is itself a safety signal. VAGUE "we are growing so priorities shift" is a red flag.
T3	Walk me through how the last few problems you worked on got chosen.	High v3	describes the provenance of recent problems, self-found or assigned, and roughly the mix.	only assignments means ticket-taking. No process at all means problems appear from nowhere, could be mixed signal either full ownership, or none at all.
T4	Walk me through the last project you owned from start to finish. Where did your part begin and end?	High v3 v6	a concrete project with the start and end of their part described.	no instance, or "I do a bit of everything", is vague, roles are not clear, expectations and evaluation are not organized.
T5	Give me an example of a fundamental problem someone on the team solved. How did that work get started?	High v3 v7	a specific fundamental problem and the origin story of the work.	only routine reporting or A/B tests means no fundamental work is being done. "We all do interesting things" is POLICY.
Colleagues and collaboration
T6	Walk me through the last real project you worked on with some other DS: who did what, and how did you decide together?	High v2	names people and a real division of labor. Very hard to fake.	"we are very collaborative" with no names or roles is BROCHURE.
T7	When you were last stuck, who did you go to, and what happened?	High v2	a named person and concrete help, including how long it took.	"I just figure it out myself" signals a silo or a competitive team, both against value 2.
T8	Give me a recent example of how work in progress gets shared on the team.	High v2	a recent concrete example of drafts and half-formed work being shared.	no example means sharing is a slogan, and outcome-only sharing is the exact failure from previous work. Be ready for no answer and keep it light.
T9	Who on the team do you learn the most from, and how does that happen?	Med v2 v6	names a mentor and what specifically they learn.	"no one in particular" means no mentorship and a growth gap.
T10	Is there anyone here you would avoid asking for help? What happens when someone's work is weak?	High v2 v7	candid about a friction point and how the team handles weak work.	"everyone here is great" read as DEFENSIVE, a signal it is not safe to criticize.
T11	How is recognition handed out, individually or as a team? When was the last time a win was celebrated?	High v2 S1	names a recent celebration and states whether ranking exists.	deflecting to "leadership recognizes us" hides individual ranking and competition.
Pace and hours
T12	When did you last work past 6pm or on a weekend? How often does that happen?	High v5	a specific dated instance plus a real frequency. This is the single best pace question.	"sometimes, it depends" with no number, or "we work hard and play hard", are the classic dodges.
T13	Tell me about the last real crunch. What caused it, and how did it end?	High v5	cause, course, and resolution, and the tone. Pride in the crunch is a red flag in itself.	"it was intense but we shipped" that romanticizes hours says the culture rewards it.
T14	We are a global team across different timezones, so messages can land outside someone's hours. Walk me through the last time that happened to you and what you did.	High v5	a specific story of a message landing outside hours and what they did, including "no, but I often answer anyway", which reveals the real expectation.	no instance, or "we are flexible", usually means always-on.
T15	How are deadlines set, and what actually happens when something slips?	Med v5	slippage is tolerated and re-planned, not punished.	"we hit our dates" hints at pressure that is never pushed back.
Psychological safety and fairness
T16	Tell me about the last time you disagreed with a senior person here. What happened?	High v7	a real episode with the disagreement and the outcome.	no episode after some tenure is a strong "not safe" signal. "I always speak up" with no story is POLICY.
T17	Tell me about a time a decision changed because some IC pushed back.	High v7	a concrete decision that flipped, and who pushed. Tests whether dissent has ever moved anything.	nothing comes to mind means dissent is performative, not effective.
T18	If you could change one thing about this team, what would it be?	High v7 v3	a specific, even mild, criticism offered without fear.	"nothing, it is perfect" is either checked out or guarded. "More headcount" is a safe non-answer.
T19	What do you wish you'd known before you joined?	High v7 v4	a mild downside or surprise, stated plainly.	"nothing" read as DEFENSIVE. "How much opportunity there is" is BROCHURE.
T20	Walk me through how the last reorg or layoff was handled: how it was decided, how it was communicated, and how the team felt afterward.	High v7 security	a full account of how it was decided, communicated, and felt. This is also where I gather leverage to negotiate a strong severance up front.	"I cannot comment" is a red flag for post-layoff resentment and thin safety.
Growth and evaluation reality
T21	Have you grown in the direction you wanted here, or the direction the company needed? Example?	High v6	open about the mix, including drift toward company needs.	"lots of growth" is VAGUE and hides whether growth points where I want.
T22	Where do people in this role go next? How many people moved up in the last year?	High v6	names destinations and a rough count over the last year.	"many directions" with no names means either no ladder or high churn.
Values and pay in practice
T23	What is something the company shipped recently that you are genuinely proud of?	Med v4	a real shipped thing, not a press release. Compare across people.	naming a generic product line is BROCHURE.
T23b	How does the work you do connect to the company's bigger goals? Give me an example.	Med v4 S1	names how their work maps to a bigger company goal, which shows whether the work matters and whether they know it.	no link means the work is disconnected from the mission, a value-4 and S1 signal.
T24	Is there anything the company does that you are not comfortable with?	High v4	names a real tension in the business.	"no, I love it here" is guarded or shows no ethical awareness, both informative.
T25	Do you feel fairly paid for what you do here? Has comp ever driven someone you know to leave?	High v1	a candid read, possibly naming a departure.	"comp is competitive" is BROCHURE and suggests soft bands.
T26	If you could change one thing about the company, what would it be?	High v4 v7	a specific change with a reason. Cross-check against teammate T18 and director D16.	"more resources" or "nothing" is a safe dodge.
Hiring manager / my future manager
Role reality and success definition
HM1	What does success look like for this role at three, six, and twelve months?	High v3 v6	specific, measurable, role-specific outcomes.	"make an impact, be a team player" hands me no definition, the exact value-3 red flag of a manager who cannot say what they need.
HM2	Walk me through what someone in this role actually worked on last month.	High v3	knows the actual day to day and describes what was worked on. I judge the exploratory share myself.	cannot describe it means they are detached from the work or the role is undefined.
HM3	What would I own end to end in my first year?	High v3 v6	a named area of ownership with an outcome.	"you will support the team" means a slice, not an arc.
Technical depth and direction
HM4	What is your own technical background, and when did you last go deep on a problem with someone who reports to you?	High v3	real depth and a specific recent example with a report.	"I am more of a people manager" is the exact non-technical-manager red flag, the condition that breaks value 3.
HM5	Describe a time a report hit a hard technical roadblock. What did you actually do?	High v3 v7	concrete, constructive engagement with the blocker.	"I trust them to handle it" means absent at the hard moment.
HM6	What is the last thing you personally learned from someone on your team?	Med v3 v7	a specific thing learned. Tests humility and two-way learning.	"I learn from everyone every day" is POLICY.
HM7	What is genuinely hard about the problems this team works on?	High v3	names the genuinely hard parts, which is the test of a technical manager.	"the problems are exciting" means they do not understand the difficulty.
How they manage and handle feedback
HM8	Tell me about a time a report disagreed with you. What did you do, and what changed?	High v7	a real instance, the outcome, and what they changed.	"it never happens" is alarming. "We have an open culture" is POLICY.
HM9	Tell me about a time a report gave you negative feedback about you/teammate/stakeholder. What did you do afterward?	High v7	a real instance, their own flaw, and what changed afterward.	"that does not really happen" means no upward safety. "I always welcome feedback" is POLICY.
HM10	When did someone last change your mind on a decision?	High v7	a specific decision that flipped.	none means rigidity or no dissent reaches them.
HM11	Describe a time you gave a report difficult negative feedback. How did it go?	Med v7	a concrete, behaved account of what was said and done.	"I give feedback straight" is a slogan, not a story.
HM12a	How are meeting norms set here, in load and in fixed times?	Med v7	a clear, specific norm for which meetings, the load, and fixed times.	"we are flexible" is POLICY with no norm behind it.
HM12b	Give me a recent example of accommodating how someone on the team works best.	High v7	a real instance of accommodating how someone on the team works best.	"we are all adults" is no answer. Asking for the instance makes a policy reply an automatic fail.
Evaluation and leveling
HM13	What is the rubric you use to evaluate IC data scientists at my level? And at the level above?	High v6 v1	named competencies and the differentiators between levels, ideally written down.	"we do not have a fixed rubric" means promotion is political and opaque. POLICY "impact, communication, collaboration" is noise.
HM13b	How was the level for this role decided, and what would the next level up look like?	High v6 v1	names how the level was set and what the next level looks like. This is my opening to negotiate title and salary.	vagueness on how the level was set means the band is soft or the level is arbitrary.
HM14	How are people ranked, individually or on team outcomes? Walk me through the last cycle.	High v2 v7	a candid account of calibration, including if ranking exists.	"we do not rank" while a Director describes calibration is a DIVERGE and a strong red flag.
HM15	What separates a strong from an exceptional IC here?	High v6	specific, observable differentiators.	"above and beyond" is VAGUE.
HM16	Tell me about a launch that failed. How was it reviewed, and what feedback did the people involved get afterward?	High v7 v6	an account of the review and the feedback the people involved got afterward, offered without prompting.	refusing to name any feedback, or an IC scapegoated for a reasonable risk, is a blame-culture signal. I ask it neutrally so I do not steer them toward the answer I want.
Growth and trajectory
HM17	What does it take to get to the next level, specifically for me?	High v6	concrete, plausible criteria for my path.	"let us focus on the role first" means no ladder or no intent to develop me. "Do great work, it will be noticed" is POLICY.
HM18	Where have the last two people in this role gone next?	High v6	named destinations.	"no one has left" means no growth or an island.
HM19	What capability would I build in year one that I do not have yet?	Med v6	a specific new capability.	"learn our stack" is not growth, it is onboarding.
HM20	Where have people in this role gone after a few years, and how does the team support the direction someone wants to grow?	High v6	names real destinations and how the team supports the direction someone wants to grow. I can read whether that path heads toward directing research.	"we will see" or no named destinations means no support and likely pull.
Pace and expectations
HM21	Walk me through how the last deadline was set. What happened when something slipped?	High v5	realistic and pushable, with a specific example of a slip handled well.	"we hit our dates" signals immovable deadlines and pressure.
HM22	Describe a time the team decided to slow down to get something right.	High v5	a real instance where deliberate pace was chosen.	cannot think of one means speed always wins.
Team, hiring, and money
HM23	How is the team structured, and where would I sit?	Med v2 S2	a clear structure and where I sit in it.	vagueness means a fluid or siloed org, or they are hiding the size.
HM24	Why is this role open, new or backfill? And what is the team's biggest gap right now?	High v2 S2 legitimacy	candid about new versus backfill and the team's real gap.	"we are growing" and nothing else raises ghost-job and backfill-after-departure questions, and why the last person left.
HM25	How is my team measured, on user benefit or on revenue? Tell me about a recent tradeoff.	High v4	a real tradeoff with a winner named.	"both, in balance" and "we are user-first" are BROCHURE and hide which side really wins.
HM26	What is the funding and revenue picture, and when did the team last see raises or promotions?	High v1	candid about financial health and the comp cycle.	"we are doing great" is BROCHURE. Vagueness about bands and raises is the compensation-opacity red flag of value 1.
Stakeholder / cross-functional partner (PM, eng lead, business, exec partner)
Decision rights and roadmap influence
ST1	Walk me through how a data-science project gets decided and scheduled here. Who is in that conversation, and who sets the deadlines and the KPIs?	High v3 v6 S1	names the decision-makers and whether data science is one of them. Ask this of everyone in every audience and compare.	cannot say means data science is a service desk. "We all collaborate" evades the power question. Divergence across audiences is a strong signal.
ST2	When data science and product disagree on what to build, who wins, and how is it resolved?	High v3	a real resolution mechanism, with an example.	"we never disagree" means data science has no standing.
ST3	Where do the problems data science works on come from? Walk me through a recent one.	High v3	names the provenance of recent problems and whether data science can originate them.	"whatever the business needs" with no example means requests only, against value 3.
ST4	Tell me about a decision data science changed that would not have happened through the normal process.	High v3 S1	a concrete, non-routine decision that data science changed.	only routine experiment outcomes means DS work is absorbed into process and rarely changes a real call, the core value-3 and S1 fear.
How DS is treated and how work lands
ST5	How do you use data-science output in your decisions? Recent example?	High S1	a concrete decision their output fed.	"we look at the dashboards" means data science is reporting, against value 3.
ST6	How do you and data science split the work when there is a decision to make? Walk me through the last one.	Med S1 v3	names the actual division of labor between their team and data science, with an example.	"true partners" with nothing behind it is BROCHURE.
ST7	What makes data science easy or hard to work with?	Med v2 v3	a candid critique of data science, which also reveals how safe they feel.	"they are great" is guarded.
Collaboration interface
ST8	How does your team work with data science day to day? Tell me about the last time it went badly.	High v2	a real friction story and how it resolved. Divergence with the manager or a teammate is a strong signal.	"it is always smooth" is DEFENSIVE.
ST9	When a model underperforms, who owns the metric, data science or you?	High v7 S1	shared ownership of outcomes.	"that is on data science" means DS is the fall guy, against value 7.
Measurement and incentives
ST10	What KPI are you judged on, and how does data science feed it?	Med v4 v3	a concrete link between the work and their scorecard.	no link means data science sits at the margin of real decisions.
ST11	Walk me through the last time two goals for the product pulled in different directions. How was it resolved?	High v4	a real tension named, with the winner, without me revealing which side I favor.	"they are aligned" or avoiding the question means the dominant side simply wins, and I learn which without signaling my preference.
Reality check
ST12	What is the biggest bottleneck in getting data-science work into your decisions?	High v3 S1	a specific blocker (trust, latency, tooling, politics).	"no bottlenecks" means no self-awareness.
Director / skip-level
Strategy and trajectory
D1	How do you decide what the team invests in learning versus just delivering?	High v6	a real policy, even a fuzzy one, that shows thought.	"we do both" means no thought about development.
D2	Where are you taking this org over the next two years? What is the bet?	High v6 S2	a specific bet, with the reasoning.	"grow and innovate" is BROCHURE.
D3	What phase does the company feel like it is in, taking new ground, scaling, or optimizing?	Med S2	a candid read of taking new ground, scaling, or optimizing. Cross-check with research.	an unclear answer means the stage read is unreliable.
D4	What broke the last time the company grew, and how did you handle it?	High S2 v7	a real failure and how it was handled.	"we scaled well" is BROCHURE.
IC evaluation and leveling
D5	What is the principal rubric for data-science evaluation across the org, and what is specific to my level?	High v6 v1	concrete, ideally documented, and differentiated by level.	"we use a standard framework" with no detail means promotion opacity. "ask your manager, there is a file for that" is disengaged. POLICY "impact and values."
D6	What does promotion to the next level actually require here?	High v6	specific criteria that match what the manager says.	a mismatch with HM17 is a DIVERGE and shows the ladder is undefined.
D7	How do you calibrate levels across the org? Recent example?	Med v1 fairness	a real calibration story.	"we do it carefully" is POLICY.
How work gets seen
D8	How does an individual's work reach leadership here? Give me the last example of IC work that reached the executive level.	High S1	the last example of IC work that reached the executive level, and how it got there.	"we have town halls" or "we celebrate wins" with no example means it does not really reach the top. Asking for the last instance keeps it non-leading.
D10	When the team decides what to work on next, walk me through the last time. Who proposed it, and what decided it?	High S1 v3	an account of the last prioritization, who proposed it, and what decided it, which shows whether IC-originated work is already weighted.	"we are very data-driven" avoids the question. No last instance means prioritization is opaque.
Resourcing and priorities
D11	How are data-science budget and headcount decided, and where does this team rank?	Med v1 v3	a candid account of how resourcing is decided.	vagueness means ad hoc resourcing and role fragility. This is flag for how layoffs will happen.
D12	When an IC proposes a problem worth solving that is not on the roadmap, what happens? Last time?	High v3	a path exists and a real instance of it being used.	"we stick to the roadmap" means ticket-taking, against values 3 and 6.
D13	What is the number one thing slowing the org down right now?	High v3 v7	a specific, candid blocker.	"nothing major" or "hiring" are dodges that reveal low openness.
Culture and safety from the top
D14	How is disagreement handled at your level? Tell me about the last time someone on your team pushed back and won.	High v7	a real instance with an outcome.	"we debate openly" with no instance is POLICY.
D15	Is there anything the company does that you are not comfortable with?	High v4	a real tension in the business, named plainly.	"nothing" means guarded or no ethical awareness.
D16	If you could change one thing about the company, what would it be?	High v4 v7	a specific change. Compare with teammate T26 and T18.	"more resources" or "move faster" are safe non-answers.
D17	How was the last layoff decided and communicated?	High security v7	a candid account of the process and how people were treated.	"that is confidential" is DEFENSIVE, and how they talk about it reveals how safe the org is.

Reverse Interview Questions

Leave a comment