The interview questions candidates score worst on are not the ones they prepare for. Real data from 816,000 sessions explains why.

Every software engineer who has spent months preparing for technical interviews knows the routine. LeetCode problems in the morning. System design walkthroughs in the evenings. Mock interviews on weekends. By the time the actual interview arrives, candidates can reverse a linked list, design a distributed cache, and explain the tradeoffs of eventual consistency without hesitation.

Then the interviewer opens with: "Can you tell me a bit about why you want to work here today?"

The room goes quiet. Not because the candidate does not know the answer, but because they never practiced it. They assumed it was a throwaway question, a 30-second warm-up before the real interview began. They treated it like small talk.

That assumption turns out to be wrong, and it costs candidates more than any failed coding problem ever does. Real session data from Final Round AI, drawn from 816,927 records across 35,511 unique interview sessions conducted between October 2022 and September 2025, shows exactly where candidates lose points. The pattern is striking: the lowest-scoring questions are almost all behavioral openers. The questions candidates skip in prep are the questions that damage their scores the most.

The scoring gap no one talks about

When the data is sorted by average candidate score, the bottom of the list is dominated by questions that most interview prep advice barely covers.

"Can you tell me how you heard about this position?" averages 25.7 out of 100 across 49 recorded sessions. That is not a rounding error. Candidates are essentially blanking on a question that sounds like it should take fifteen seconds to answer.

"Why did you choose that major and school?" averages 38.5 out of 100 across 161 sessions. This is a common question in recruiting screens and early-round interviews, particularly for candidates earlier in their careers. Yet more than 160 recorded sessions show candidates averaging below 40 on it.

"Why are you interviewing here today?" averages 44.4 out of 100 across 91 sessions. And "Why do you think we should hire you?" averages 46.7 out of 100 across 42 sessions.

All four of these questions share something: they are open-ended, they require self-knowledge, and they reward structured multi-part answers rather than a single correct response. None of them appear on any LeetCode list. None of them require memorizing an algorithm. And yet, across thousands of real interviews, they produce some of the lowest scores in the entire dataset.

The surprise in this data is not that candidates struggle with hard questions. It is that they struggle most with questions designed to be easy. These questions have no hidden algorithmic complexity. They have no trick. They are simply asking the candidate to articulate something they should know about themselves. The scores suggest that most candidates cannot do this under pressure in a structured way, because they have never tried.

The role breakdown reveals a counterintuitive gap

Across 35,511 sessions, the average score by role shows meaningful differences that carry practical implications.

Product Managers score highest at 58.9 out of 100 across 2,303 sessions. Data Scientists average 58.6 across 3,360 sessions. Software Engineers average 55.0 across 10,700 sessions. Software Developers average 52.0 across 2,450 sessions.

The most counterintuitive finding: Senior Software Engineers average 52.4 out of 100 across 2,688 sessions. That puts them below regular Software Engineers at 55.0 and only marginally above Software Developers at 52.0.

This gap runs counter to the obvious assumption that more experienced candidates would perform better. There are a few likely reasons for it.

Senior candidates tend to have strong opinions and well-formed habits. That works well in technical discussions and system design. It creates problems when behavioral questions require them to adapt their communication style to what an interviewer is looking for rather than what they prefer to say. A senior engineer who has been working at one company for six years often struggles to explain their motivations in a way that resonates with a hiring team at a different company with different values and a different culture.

Senior candidates also tend to underprepare behavioral responses specifically. They have been through many interviews, they know their technical material cold, and they assume the softer parts of the interview will take care of themselves. The data suggests they do not.

Product Managers scoring highest likely reflects the nature of PM interview training. PM interview prep culture heavily emphasizes storytelling, structured answers, and behavioral frameworks like STAR (Situation, Task, Action, Result). That skill transfers directly to the kinds of questions where candidates in other roles score poorly. PMs practice saying "why I want this role" and "why I'm the right fit" repeatedly, because those questions are core to PM interviews. Engineers rarely do.

Three reasons behavioral openers score low

The pattern has structural causes beyond individual candidate preparation.

Candidates treat them as warm-up. Most interview prep advice, both formal and informal, focuses on technical questions, case studies, and behavioral questions about past performance ("Tell me about a time when..."). Questions like "Why do you want to work here?" get less than five minutes of preparation time because candidates do not believe they matter much. When the question arrives in an actual interview, it lands with the full weight of stakes but none of the preparation. Candidates improvise. Improvised answers to open-ended questions tend to be vague, which explains the scores.

Generic answers do not demonstrate fit. "I've heard great things about the company culture" or "I'm excited about your product roadmap" are answers that could apply to any employer. Interviewers evaluating these answers know when they are hearing something rehearsed and hollow. The score reflects that knowledge. A specific answer tied to the candidate's actual background, with named products or named challenges or named people, performs better. Most candidates do not prepare specific answers because specific preparation requires research, and candidates generally prioritize technical prep over company research.

Short answers to questions that reward length perform poorly. "Why should we hire you?" is not a yes/no question and it is not a one-sentence question. A strong answer has multiple parts: skills, specific examples, and a connection between what the candidate offers and what the role requires. Candidates who give 30-second answers to this question tend to leave out at least one of these components. Incomplete answers score lower, and because candidates do not treat this question as a high-stakes item, they do not invest the time to build a complete answer. The result is a score that looks like a candidate who was not paying attention, even when the candidate is otherwise technically strong.

What to do with this by role

Software Engineers (average: 55.0) should audit which question types are missing from their prep. Most SE candidates have a solid technical foundation but have never run a timed mock for "Why do you want to work here?" or "Why should we hire you?". Building one or two strong answers for motivation questions, practiced out loud rather than just thought through, closes most of the gap. The goal is not memorization. It is having a structure: one sentence on personal motivation, one sentence on the specific role, one sentence on the company specifically. That three-part structure takes 60 to 90 seconds to deliver and scores materially higher than an improvised paragraph.

Product Managers (average: 58.9) already outperform other roles on behavioral questions, but they are not immune. PM candidates who score in the lower range tend to do so on quantitative or technical questions where they do not have strong frameworks. The tradeoff is that their behavioral score is strong. For PM candidates, the priority is maintaining that strength while not neglecting any technical components that appear in specific company interview processes. The data shows PMs are doing something right on behavioral prep. The lesson is to be intentional about what that is, rather than assuming it will carry over from job to job.

Senior Software Engineers (average: 52.4, below regular SWEs) have the clearest improvement path. The gap is behavioral, not technical. Senior candidates should practice articulating transitions: why they are leaving their current role, what they are looking for specifically, and how this particular opportunity fits. Vague answers to these questions are what drag the score down. A senior candidate who can clearly explain "I want to move from infrastructure work into product-facing systems because..." and follow it with concrete evidence will outperform the average by a meaningful margin. The fix is not hard. It requires a couple hours of deliberate prep on questions senior candidates have avoided for years.

For all roles, one practice pattern helps more than most: run a full mock that starts with the behavioral opener rather than the technical section. Most candidates warm up by practicing coding or case questions, which means the first time they say "Why do you want to work here?" under any kind of pressure is in the actual interview. Running the warm-up questions first, out loud, with a timer, produces noticeably better real-interview performance.

Where the full data lives

The analysis above draws on a subset of the findings. Final Round AI's full breakdown, which includes charts showing the role-by-role score distribution and a deeper breakdown of the specific behavioral question categories where scores diverge, is at https://www.finalroundai.com/blog/interview-questions-candidates-struggle-with.

The underlying dataset covers sessions across hundreds of companies and multiple years, making the pattern reliable rather than a result of any single industry or time period. The core finding holds across cohorts: candidates score lowest where they prepare least, and they prepare least for the questions that feel too easy to merit practice.