Goldman Sachs Investment Banking is supposed to be the hardest finance interview. The data says something different.
Final Round AI analyzed 9,494 live interview sessions across JP Morgan, Goldman Sachs, Capital One, Wells Fargo, Bank of America, and American Express — all captured through their Interview Copilot tool during actual job interviews, not practice runs. The results challenge most of the conventional wisdom about which finance company is hardest to interview at.
The counterintuitive finding: Wells Fargo outscores Goldman Sachs
The full company ranking by average answer quality score (0-100 scale):
- Wells Fargo: 58.1/100 across 1,498 sessions
- Capital One: 56.5/100 across 1,883 sessions
- Goldman Sachs: 56.3/100 across 1,423 sessions
- American Express: 56.2/100 across 1,309 sessions
- JP Morgan: 55.9/100 across 1,981 sessions
- Bank of America: 55.7/100 across 1,400 sessions
Wells Fargo, often treated as second-tier, produces the highest candidate answer scores in the dataset. Goldman Sachs sits third. JP Morgan, the largest bank by session count, ranks fifth.
The scoring model measures answer completeness and structure, not interview outcomes. It rewards STAR-format answers and penalizes vague or incomplete ones. For context: Amazon averages 57.5 and Google 56.8 in the same dataset. Finance and tech interviews are essentially equivalent. Treating finance as an easier track than a Google loop is not supported by this data.
The Goldman Sachs IB finding is the real story
Goldman Sachs Investment Banking Associate candidates averaged 39.7 out of 100 across 308 sessions. That is 16.4 points below Wells Fargo company-level average and 27.9 points below Goldman own Technology Cyber Risk Advisory role (67.6 across 161 sessions).
That gap is not about candidate quality. It is about question type.
When a Goldman IB interviewer asks you to walk through a leveraged buyout, the correct answer is two to three sentences: entry multiple, debt structure, exit assumptions. Not a narrative. The scoring model Final Round AI uses is calibrated for extended, structured responses. Brief, technically precise answers score low on that model regardless of accuracy. Candidates preparing for Goldman IB with general behavioral coaching are optimizing for the wrong format.
Role breakdown:
- Goldman Sachs IB Associate: 39.7/100 (308 sessions)
- Goldman Sachs Technology Cyber Risk Advisory: 67.6/100 (161 sessions)
- JP Morgan Software Engineer: 56.2/100 (280 sessions)
- JP Morgan Associate: 51.2/100 (119 sessions)
- American Express Machine Learning Engineer: 71.0/100 (147 sessions)
- Bank of America Middleware Engineer: 45.8/100 (112 sessions)
The spread within companies is wider than the spread across companies. The role matters more than the firm when calibrating preparation.
What changed from 2024 to 2025
Capital One improved 3.5 points (55.7 to 59.2). American Express improved 4.9 points (54.8 to 59.7). Both moved upward across 3,192 combined sessions.
Wells Fargo dropped 5.8 points (59.1 to 53.3) across 1,498 sessions. The largest single-year swing in either direction. JP Morgan declined 1.9 points. Goldman was flat.
If you are interviewing at Wells Fargo this year, do not calibrate from 2024 data. The 2025 figure of 53.3 puts Wells Fargo at the bottom of all six companies for that year.
What this means in practice
For Goldman Sachs IB Associate candidates: behavioral coaching is not enough. Build a parallel technical track. Practice explaining a three-statement model in under 90 seconds. The format that works for tech and risk roles actively hurts performance in IB-specific questions.
For Capital One Data Engineer candidates (196 sessions, avg 61.1): the improving trend and above-average scores validate the technical-plus-behavioral format. SQL, system design, and STAR behavioral prep is the right combination.
For Bank of America Middleware Engineer candidates (112 sessions, avg 45.8): this role scores 10 points below the BofA company average. Practice out loud. Push for specific, outcome-driven answers.
For Wells Fargo candidates in 2025: use 53.3, not 58.1.
Final Round AI full report with difficulty ranking chart and year-over-year trend data: https://www.finalroundai.com/blog/finance-company-interview-difficulty-ranking
Top comments (0)