Investment Banking HireVue Questions 2026: Worked Answers

What the banks publish about the investment banking HireVue round

Two banks describe this stage on their own pages, and they describe different parts of it. Reading both beats reading either.

Goldman Sachs is the only bank found that names the vendor. Its student preparation page writes "if selected following your HireVue interview, you'll be invited to a series of final round interviews," which settles both the platform and what sits on the other side of it. The page returned 403 to a direct request and was read with browser headers, checked 30 September 2026.

Deutsche Bank never names a vendor. Its application FAQ says "the technology we use will record you through the platform," and adds that "you will have the opportunity to test it first and complete an example question." So the stage exists, the warm up exists, and the vendor stays unnamed.

One caution about a third bank. Morgan Stanley writes that "interviews could be competency, strengths based, skills based or technical." That sentence covers its interviews in general. Nothing on the page attaches those families to a recorded round, so treat it as evidence about the bank rather than about this stage.

The timing on a HireVue question, and who sets the numbers

Deutsche Bank publishes the strictest version of this round in print: "after you see each question you will be provided with two minutes to prepare your response. You then have two minutes to record your responses and just have the one attempt."

Compare that with what the publisher recommends to its customers. HireVue's explainability statement lists preparation time as a setting from zero to five minutes, recommends 60 seconds, and recommends "1-2 retries in addition to the first attempt."

Both are true at once. The employer buys the assessment and picks the values, and this bank picked no retry at all. Any page that promises you a redo is describing somebody else's configuration.

The arithmetic of the sitting follows from those two numbers. Four minutes per question means a five question interview occupies twenty minutes of clock, and the publisher puts total candidate time at 15 to 25 minutes for three to six video questions plus two to three games. Goldman's 30 minutes sits just above that, which is consistent with a longer answer cap or a longer set.

Two working rules come out of this. Plan every answer as a single take, because one bank in print gives you exactly one. And read the instruction screen on the day, since it carries your actual preparation window and your actual attempt count. For the band your recording lands in and the five word floor below which nothing is scored, the publisher's own mechanics are set out where that document is read in full.

Worked answer one: why this division, in four moves

This is the question banks put in writing. Deutsche Bank expects you to explain "why the division you're applying to is more suitable for you than another." Morgan Stanley asks for "a compelling reason for why you want to work in this industry, our Firm, and the division you're interested in."

So the answer needs a reason that could only be about that division. Here is a two minute answer of ours, for a question asking why leveraged finance.

"I want leveraged finance because I like the part of the job where the capital structure decides whether a deal happens. I spent last summer in a treasury team at a mid sized manufacturer, building the covenant schedule for a refinancing of about 150 million dollars. The interesting work was the sensitivity table, where a one turn move in the leverage multiple changed what the sponsor could pay. That is the daily work in this division, and it is why I prefer it to an M and A seat. The risk in my profile is that my modeling is self taught rather than trained, so I have been rebuilding the same leveraged buyout model weekly to close that gap."

Four things carry that answer, and they are the dimensions our own synthesis rubric scores, read from the code that does the scoring.

Those weights are ours. No bank publishes a rubric for this round, and the publisher publishes bands rather than criteria.

Worked answer two: the behavioral item, with a result you can count

Deutsche Bank runs a "Behavioral Questionnaire" before the recorded round, and the recorded questions in this family ask for a past episode. What banks publish about that half is thinner than what consulting firms publish, and the gap is the preparation problem. Take a common one: tell us about a time you found an error in someone else's work.

A weak version stays in the first person plural and never lands. "We were working on a valuation and I noticed some of the inputs looked off, so I flagged it and we corrected the model together, which taught me the importance of attention to detail."

A stronger version of ours, still inside two minutes: "During a student fund pitch, our target's revenue line grew at 12 percent while the company's last filing showed 4 percent. I checked the source and found we had pulled a consensus estimate into a historical column. Correcting it cut our fair value by about a fifth, which turned a buy into a hold. I told the team lead before the pitch rather than after, and we presented the hold with the correction on the slide. We now keep a one page source list for every input."

The difference is scoreable from the transcript alone. There is a situation, a number that came from a document, an action taken by the speaker, a consequence that changed a decision, and a change that outlasted the episode. The publisher states that its model "relies only on what is said by the candidate," so anything you intend to be credited has to be spoken.

One structural habit helps here. Say the outcome before the clock gets tight, then use the remaining seconds for the detail. An answer that runs out mid story loses the only sentence that earns the band.

Worked answer three: a situational item answered with a decision

Situational questions ask what you would do. They are the family Deutsche Bank tests separately with a situational judgment test, and the publisher lists "situational judgement questions" among the shapes its customers use.

Our worked item: you are an analyst, it is 7 p.m., and you have been asked for a comparable companies page by 9 p.m. for a client meeting tomorrow. You also owe a vice president a model review tonight. Both people are senior to you.

A model answer of ours, which is a decision rather than a description: "I would send the comparables page first, because it has an external deadline attached to a client meeting and the model review does not. Before starting I would send the vice president two lines: the page is due at nine, I expect to be on the review by nine fifteen, and if that is too late I need to know now. Then I would cut scope rather than quality on the page, so six comparables I can source properly instead of ten I cannot. At nine I would flag anything I had to leave blank in the email rather than hiding it."

The moves that earn the score are the order, the escalation before the crunch, the explicit scope cut, and the disclosure of what is missing. Notice that the answer takes a side in the first sentence. A recording that lists considerations and never chooses gives a rating scale nothing to anchor on, and these questions are built so that a choice exists.

A word on tone, since this is the family candidates over polish. The publisher's own tips page says "don't worry about making eye contact with the camera," checked 29 September 2026. What the platform does record, and what it does not, is fact checked separately.

Worked answer four: the light technical item, done out loud

Here is where honesty matters more than coverage. No bank read for this page states that a technical question appears in its recorded round. Goldman gives the duration and nothing about content. Deutsche Bank gives the timing rule and the motivation themes. Morgan Stanley names technical interviews without tying them to a video stage. Prepare one anyway, because the cost of being ready is twenty minutes, and treat every page that promises you a specific technical question as unsourced.

A worked item of ours, of the size that fits a two minute recording. A company reports EBITDA of 80 million dollars and net debt of 240 million dollars. A close peer trades at 9 times EBITDA. What is the implied equity value?

Say the chain out loud in this order.

1. Name the bridge first. Equity value equals enterprise value minus net debt, for a company with no minority interests or other adjustments.
2. Apply the multiple. 9 times 80 million dollars is 720 million dollars of enterprise value.
3. Subtract the debt. 720 million dollars minus 240 million dollars leaves 480 million dollars of implied equity value.
4. Say what would change it. The peer multiple is one company, so a range of 8 to 10 times would put equity value between 400 and 560 million dollars.

Two habits inside that answer are worth more than the number. The unit is stated at every step, which is where spoken math usually breaks. And the last line converts a point estimate into a range, which is the move that reads as trained rather than rehearsed.

Worked task five: the game section, and one numeracy level

Some banking assessments include the publisher's games alongside the recorded questions. HireVue describes them as short tasks that give "insight into a range of cognitive skills, including numeracy, problem-solving, and attention," and states that they adapt: clear a task and the next is harder, fail one and the next is easier.

We cannot reproduce a real level, and we would not. Here is an item of ours built to the published shape, the kind of comparison a numeracy task asks you to make under a few seconds of clock.

Two offers sit on screen. Offer A gives 40 percent of 250. Offer B gives 15 percent of 700. Which is larger?

40 percent of 250 is 100. 15 percent of 700 is 105. Offer B is larger, by 5.

The method that survives the clock is order of magnitude before precision. Both products land near one hundred, so the decision needs the second digit, and the fastest route is to take ten percent of each base first: 25 and 70. Four lots of 25 make 100, and one and a half lots of 70 make 105. Because the levels get harder as you clear them, expect to fail one near the end. That is the design working as published rather than a signal about you.

What no bank publishes about this round

Four gaps, and naming them is more useful than filling them.

The question count. None of the four bank pages read on 30 September 2026 gives one. The publisher's ranges of three to six video questions, or five to eight in its candidate FAQ, are product ranges rather than facts about a given bank.

Whether a technical question appears. Covered above. It is the single most confidently repeated claim about this round and the one with no employer source behind it.

Who watches your recording, and when. The publisher says the final hiring decision belongs to the employer and that low band interviews are viewed when volumes are short. No bank states whether a human watches before a rejection.

The bar. Nobody publishes a cutoff, and the same platform serves different roles at different bars inside one bank. Any page quoting a score you need for a named bank has invented it.

One rule the banks do publish, and it is worth obeying. UBS writes that "during recorded or virtual video interviews, your answers must be your own. Using AI to generate responses is strictly prohibited." Deutsche Bank uses softer words for the same rule, saying it strongly recommends against AI tools because it is "looking for authenticity in your responses."

Rehearsing for a single take, and where to drill

The plan below is indexed on what you find when you record yourself, rather than on a calendar.

1. Write six answer skeletons, one per family. Motivation for the industry, motivation for the division, one behavioral, one situational, one light technical, one on a market you follow. The publisher tells candidates "you can even keep your notes next to you," so these can sit beside the laptop as titles.
2. Record each one once, then watch it with the sound on. Whatever you do on the second take you will not get at Deutsche Bank, so the first take is the artifact to judge.
3. Label what broke, then fix that one thing. If the answer ran long, cut the setup. If it had no number, add the number. If it never chose, put the choice in the first sentence.
4. Rehearse the division answer last and aloud. It is the one both banks put in writing, and it is the one that sounds rehearsed fastest when it is read.

A time budget helps with a two minute cap. Fifteen seconds for the answer itself, sixty for two supports, twenty for the risk or the caveat, fifteen to close. That is 110 seconds, leaving ten for the pause you will take when you start.

Our free drills score the transferable part of this, which is leading with a conclusion and supporting it under a clock, and how an answer is scored publishes the weights they run on, read from the code. What the drills do not do is record video, score speech, or reproduce a bank's question set. Our assessment simulators cover firm built formats, and none of the six is live yet, which that page says itself. For the publisher's mechanics in full, read the platform guide. Everything above is either a bank page or a publisher document, dated in the body, or is labeled as ours.