The Complete Guide to Investment Banking Simulations: What Top Firms Actually Test

For its 2025 internship class Goldman Sachs received more than 360,000 applicants and accepted fewer than 1% of them: a 15% rise on 2024 and, on the bank's own account, up 300% since 2018. Source: Fortune, "A record 360,000 students applied for a Goldman Sachs internship this year, and fewer than 1% were accepted", 16 June 2025. Yet the skills tested in IB simulations are more accessible than most candidates believe.

What IB Simulations Actually Measure

Based on our analysis of simulation programs offered by Goldman Sachs, J.P. Morgan, and Morgan Stanley, investment banking simulations consistently assess five core competency areas:

1. Financial Statement Analysis: Reading and interpreting income statements, balance sheets, and cash flow statements to identify key drivers of business performance.
2. Valuation Methodology: Building discounted cash flow (DCF) models, comparable company analysis (comps), and precedent transaction analysis.
3. Deal Structuring: Evaluating M&A scenarios, understanding debt vs. equity financing, and assessing strategic rationale.
4. Client Communication: Synthesizing complex financial data into clear, compelling narratives for pitch books and client presentations.
5. Attention to Detail Under Pressure: Performing accurately when working with large datasets and tight deadlines.

The Numbers Behind IB Recruiting

Selectivity, not pay, is the number this page can stand behind: no compensation figure for first-year analysts in New York or Paris is printed by any regulator, filing or dated survey this desk could open. What is on the record is the ratio. At its 2023 Investor Day Goldman Sachs disclosed roughly 300 applicants for every open position. Source: eFinancialCareers, "Goldman Sachs has 300 applicants per open position", February 2023.

Breaking Down a Typical IB Simulation

A standard investment banking simulation runs 60-120 minutes and typically follows this structure:

Phase 1: Market Research (15-20 min): You receive a brief on a fictional or disguised real company and its industry. You're expected to identify market trends, competitive positioning, and growth drivers.

Phase 2: Financial Analysis (25-35 min): Using provided financial statements, you build a basic valuation model. This tests your ability to calculate Enterprise Value, EBITDA multiples, and free cash flow.

Phase 3: Strategic Recommendation (15-25 min): Based on your analysis, you prepare a recommendation memo or mini-pitch deck. Evaluators assess your logic, clarity, and ability to defend assumptions.

Phase 4: Q&A or Peer Discussion (10-15 min): Some simulations include a discussion component where you defend your analysis or collaborate with other candidates.

How to Prepare

No published study this desk could open quantifies how much rehearsal raises a simulation score. What the evidence does support is how much weight employers now put on assessment itself: 56% of employers use pre-employment assessments to gauge applicants' knowledge, skills and abilities, and 79%³ say skills assessments are just as or more important than other hiring criteria: across all industries, not banking alone. Source: SHRM, "New SHRM Research Makes the Case for Skills-Based Hiring", 15 August 2022. The key preparation areas are:

• A modelling test exercises model architecture before it exercises anything else.
• Accounting Fundamentals: Understand how the three financial statements link together. Every modelling exercise rests on it.
• Industry Knowledge: Read the Financial Times, Bloomberg, and Wall Street Journal daily. Understanding current deal flow gives you context for simulation scenarios.
• A recommendation that cannot be stated in the client's language does not survive the Q&A, yet

The Democratization Effect

What a banking simulation is really testing

Candidates prepare for banking assessments as if they were exams in valuation. They are not. A simulation is designed to observe whether someone can hold accuracy, speed and composure simultaneously, because the failure mode the desk fears is not ignorance, it is a confident number that is wrong under time pressure.

Concretely, four competencies are being separated:

1. Numerical discipline. Not mental arithmetic, but whether the candidate sanity-checks an output before presenting it. A margin that implies a negative cost base should stop the candidate mid-sentence.
2. Structuring under incompleteness. The pack deliberately omits something material. Assessors watch whether the candidate flags the gap and proceeds with a stated assumption, or silently invents a figure.
3. Commercial reasoning. Whether the recommendation is expressed in the client's terms (dilution, covenant headroom, control, timing) rather than in the model's terms.
4. Composure under challenge. A managing director will push on a number that is correct. Conceding a right answer under social pressure is a disqualifying signal; defending it rudely is a different one.

The modelling test, decomposed

The modelling component is where preparation pays most predictably, because the rubric is close to mechanical. Assessors typically check: whether the model's assumptions live in one clearly marked block rather than being hardcoded into formulas; whether the historical period ties to the source statements; whether the sensitivity is genuine (does the output actually move with the driver); and whether the candidate can state, without scrolling, which two assumptions drive most of the answer.

Two habits distinguish candidates who finish. First, build the smallest model that answers the question and extend only if time remains: a half-finished three-statement build scores below a complete simplified one. Second, reserve the final five minutes for a check pass rather than a formatting pass. Panels report that the most common fatal defect is an arithmetic error the candidate had time to catch and never looked for.

The case and the client conversation

The second half of most assessment centres is a client-facing exercise: a board is considering a disposal, a founder is weighing a minority sale, a corporate wants to fund an acquisition. Here the model is a supporting document and the judgement is the product.

Strong candidates do three things. They state the client's objective before any option ("you are optimising for control retention, not headline value") because a recommendation without an objective cannot be evaluated. They present two viable routes with the trade-off named, rather than one preferred answer, since real mandates are chosen against constraints the adviser does not control. And they say explicitly what they would need to verify before committing: diligence gaps, financing conditions, timing risk. That last habit reads as maturity; its absence reads as someone who has never been wrong in front of a client.

A structured eight-week preparation plan

1. Weeks 1–2: mechanics. Build three models from raw filings, no templates. Accuracy first, speed later. Keep an error log: the same three mistakes will recur, and the log is what removes them.
2. Weeks 3–4: compression. Rebuild each model in half the time, then again in a quarter. This is where sanity-checking becomes automatic rather than aspirational.
3. Weeks 5–6: commercial framing. For each model, write a one-page recommendation in the client's language, with two routes and a named trade-off. Have someone who is not a banker read it and tell you what the recommendation was.
4. Week 7: adversarial rehearsal. Present under challenge, including at least one push-back on a number that is correct. Practise the sentence "that figure holds, and here is why" out loud.
5. Week 8: full simulation. One end-to-end run under real time constraints, with no reference materials you would not have on the day.

What it means for candidates

The move to simulations has narrowed the advantage that target-school candidates held through networks and interview coaching, and widened the advantage held by anyone willing to do repeated, timed, reviewed reps on real filings. A simulation cannot see where you studied; it sees whether your model ties, whether you flagged the gap in the pack, and whether you held a correct number under pressure. Those three behaviours are trainable in eight weeks, and they are the same three behaviours the job rewards in the first year, which is precisely why the format was adopted.

Withdrawn figures

On 8 September 2026 this page was audited against its own reference list, document by document. The figures withdrawn from it are the United States $110,000 base and $75,000–$125,000 bonus range and the Paris €55,000–€65,000 equivalent, Dartmouth Tuck's 42% non-target share and its 15-point rise, the Corporate Finance Institute's 34% score uplift, Bloomberg's ranking of Excel, the claim that statement linkage is tested in 94% of simulations, McKinsey's 35% evaluation weighting, and the Financial Services Skills Commission's 2.8x interview multiple. No primary document published by a named body could be found printing them at any edition a reader can open, so they have been withdrawn rather than left standing on a reference nobody can check. Every figure that remains above is credited in its own sentence to a document a reader can open at a stated date.