In almost every other industry a case covers, the client knows the cost of what it sold on the day it sold it. In financial services it does not: the loan's losses, the policy's claims, and the fund's returns arrive one to ten years after the revenue is booked. That is what a financial services case is really testing, and the candidate who prices the cost that has not shown up yet, then checks the capital cushion the regulator makes the client hold against it, is the one whose scoring sheet fills up.
Most candidates walk in with a profitability tree and treat a bank like a grocer with unusually boring shelves. Revenue up, costs flat, recommend growth. In a typical first round of eight, two or three instead ask when the losses land, what funds the growth, and how much capital it eats. I remember those candidates by name.
What follows is the capital arithmetic most guides skip and a worked case you can reproduce on one sheet of paper. Corporate NPV lives in the finance case guide; payments and neobanks live in the fintech guide.
Three businesses, one accounting problem
"Financial services" in a prompt means one of three things, each with a different gap between the sale and the bill.
| Client | What it sells | Raw material | When the true cost is known | Ratios to own |
|---|---|---|---|---|
| Bank | Money now, repaid later | Deposits and wholesale funding | 12 to 36 months after origination, as the vintage seasons | Net interest margin, net charge-off rate, cost-to-income, common equity ratio |
| Insurer | A promise to pay if something happens | Premium held as float | Months (motor) to decades (liability, life) after the policy is written | Loss ratio, expense ratio, combined ratio, solvency ratio |
| Asset manager | Management of other people's money | Talent and distribution | Costs are known; revenue floats with markets | Fee rate on assets, net flows versus market effect, operating margin |
The bank and the insurer share the problem head-on: revenue is recognized before the cost of goods is known. The asset manager is the mirror image: costs are mostly salaries, fixed in January, while revenue is a fee on assets whose value the market decides in December. Either way, the tree you memorized has a branch that moves by itself. A useful habit: when an exhibit shows a year-one profit on a new product, write "cost pending" next to it before you speak.
Where the sheet gives and takes points in the first five minutes
Whatever the template calls them, interviewers score the same things in the opening minutes. What I hear, and what goes on the sheet:
| Candidate says | What I write |
|---|---|
| "Let me break profit into revenue and cost, and revenue into price and volume." | Generic tree. Neutral. Waiting. |
| "Is this a lender, an insurer, or a fee business, and is the objective profit, return on equity, or a capital ratio?" | Knows the three species. Plus one. |
| "For any growth we find, I want to know what funds it and what capital it consumes." | The balance sheet is the product. Plus two. |
| "What loss rate is baked into this year's result, and how old are the loans it came from?" | The timing problem. Plus two. Two in eight ask this. |
| "Can we raise prices?" | In a rate-driven market, price is mostly set for you. Minus one as a first idea. |
You are rewarded for naming the constraint (funding, capital, credit risk timing) before you name the lever. Levers are cheap here; every bank can lend more tomorrow. Constraints are where the case lives.
The regulator is the third person in the room
The client, you, and a regulator who never speaks but sets the rules. For banks, the Basel framework requires common equity of at least 4.5% of risk-weighted assets plus a 2.5% conservation buffer, also in common equity; national add-ons push real targets well above 7%. For European insurers, Solvency II calibrates required capital as the value at risk of the insurer's own funds at a 99.5% confidence level over one year, enough to survive all but a one-in-200-year loss.
What that does to a growth case is the point: every dollar of new lending or premium must first be matched with equity that could otherwise have gone back to shareholders, and that equity has a price.
The arithmetic takes one line. Take a $1.0bn loan book that the rulebook weights at 100% (the case gives you the weight; do not invent one). A bank targeting a 10.5% common equity ratio must hold $105m against it. If shareholders expect 15% after tax on that equity, the book must earn $15.75m after tax, or $21m before tax at a 25% rate, before it has created anything. That is a 2.1% pre-tax hurdle margin:
hurdle margin = risk weight × target capital ratio × required return / (1 − tax rate) = 1.00 × 0.105 × 0.15 / 0.75 = 2.1%
Hold every spread the case shows you against that number, not against zero. A loan earning 1.5% after funding, losses, and cost is profitable on an income statement and value-destroying to the owner. Candidates who say so out loud are rare enough that I still notice.
Worked case: Cedar Valley Bank buys growth with next year's losses
Cedar Valley is invented; all figures are case inputs.
Prompt. Cedar Valley runs a $4.0bn indirect auto loan book: 7.5% yield, 3.5% marginal funding cost, 1.0% net charge-offs, 1.5% servicing cost, equity allocated at 10% of balances. The head of consumer lending has added $1.0bn of near-prime loans (scores 620 to 659) at a 9.5% yield, funded the same way. Year one is in and the board is delighted. Should they expand the program?
Step one: the base. Pre-tax margin = 7.5 − 3.5 − 1.0 − 1.5 = 1.5% of $4.0bn = $60m on $400m of equity: a 15% pre-tax return.
Step two: year one of the new book. Auto loans rarely default in their first months; losses typically peak in the second year. Year-one charge-offs came in at 0.8%. Margin = 9.5 − 3.5 − 0.8 − 1.5 = 3.7%, or $37m.
| Year one | Base book | New book | Combined |
|---|---|---|---|
| Balances | $4.0bn | $1.0bn | $5.0bn |
| Pre-tax profit | $60m | $37m | $97m |
| Equity allocated | $400m | $100m | $500m |
| Pre-tax return on equity | 15.0% | 37.0% | 19.4% |
This is the moment the case is testing. The cost of goods on that $1.0bn has not arrived.
Step three: year two, when the vintage seasons. The interviewer hands you an exhibit: charge-offs on the 620 to 659 vintage have risen to 4.8%. Margin = 9.5 − 3.5 − 4.8 − 1.5 = −0.3%, a $3m loss.
| Year two | Base book | New book | Combined |
|---|---|---|---|
| Pre-tax profit | $60m | −$3m | $57m |
| Equity allocated | $400m | $100m | $500m |
| Pre-tax return on equity | 15.0% | −3.0% | 11.4% |
Return on equity is now 11.4%, below where the bank started, on a larger and riskier balance sheet. Year-one earnings were a forecast wearing a P&L's clothes.
Step four: the number the interviewer wants. Do not argue about whether 4.8% is typical; find the break-even. To match the bank's 15% return, the new book must earn $15m on $100m of equity, a 1.5% margin, which leaves room for a loss rate of 9.5 − 3.5 − 1.5 − 1.5 = 3.0%. The whole decision reduces to one question: is the through-the-cycle charge-off rate on a 620 to 659 auto vintage above or below 3.0%?
Recommendation, as said aloud. "Do not expand yet. Year one tells us nothing, because auto losses peak in year two. The program only works if lifetime charge-offs on this score band stay under 3.0%, and the year-two reading is already 4.8%. Cap the book at $1.0bn, price the next tranche at least 180 basis points higher, and stop originations if any new vintage's 12-month charge-off rate crosses 3.0%. Two risks: collections costs on near-prime will run above the 1.5% we assumed, and a fall in used-car prices raises the loss on every default at once."
Ned's rule. In a financial services case, a number is not a profit until you know the year its cost arrives. Ask for the vintage before you applaud the result.
The same trap in an insurer's suit
Swap loans for policies and the logic survives. Harborline Mutual, also invented, earns $500m of premium at a 65% loss ratio and a 32% expense ratio: a 97% combined ratio and a $15m underwriting profit. Float of 1.5 times premium, $750m, invested at 4.0%, adds $30m. Pre-tax profit: $45m.
The growth plan cuts rates 8% to grow premium 20%, to $600m. Expected claims per policy do not fall with the price, so the loss ratio rises to 65 / 0.92 = 70.7%. Fixed costs spread over more premium; call the expense ratio 30%. Combined ratio 100.7%, an underwriting loss of about $4m. Float grows to $900m and earns $36m. Pre-tax profit: $32m, down from $45m.
Getting back to $45m would need $49m of investment income on $900m, a 5.4% yield in a 4.0% market. Said plainly: "The plan is a bet that our portfolio managers beat their own yield by 140 basis points, permanently, to pay for a price cut. I would not make it." And the claims on this year's cheaper policies will be paid over several years, which is exactly why Solvency II sizes capital against a one-in-200-year loss rather than last year's average.
Funding is a customer decision, not a line item
Candidates trained on retail cases treat cost of goods as something you negotiate with suppliers. A bank's main input is money borrowed from customers who can take it back. Silicon Valley Bank grew from $71 billion to over $211 billion in assets between 2019 and 2021; on March 9, 2023, depositors withdrew over $40 billion in a day, and management expected $100 billion more the next morning. The Federal Reserve's review blames a concentrated depositor base, heavy reliance on uninsured deposits, and interest-rate risk that management chose to measure differently rather than reduce. For a case, the lesson is narrower: when a growth plan appears, ask what funds it and how sticky that funding is.
Mix also explains a fact most candidates get backwards: in banking, scale does not buy margin. The FDIC's Quarterly Banking Profile for the second quarter of 2026 puts the industry's net interest margin at 3.32% across 4,238 institutions, and the breakdown by size runs the wrong way for anyone expecting economies of scale.
| Asset size, second quarter 2026 | Net interest margin |
|---|---|
| Over $250bn | 2.94% |
| $10bn to $250bn | 3.95% |
| $1bn to $10bn | 3.93% |
| $100m to $1bn | 4.02% |
| Under $100m | 3.99% |
| All FDIC-insured institutions | 3.32% |
Source: FDIC Quarterly Banking Profile, second quarter 2026. The largest banks hold more securities, use more wholesale funding, and do more low-yield corporate lending, offset by fee income smaller banks lack. If an exhibit shows a bank's margin below peers, the first hypothesis is mix, not incompetence.
What the firms say, and what is folklore
Firms publish less about financial services cases than prep sites imply.
Stated. Oliver Wyman, whose financial services practice is large enough that many candidates assume every case will be a bank, describes its cases as "often based on real-world challenges" and asks for six things: pinpoint the issue, break it down, pick an approach, analyze, use numbers carefully, recommend. Its Brazil and Mexico campus page specifies a 26-question, 50-minute online test, a 15-minute fit interview, and two rounds of two case interviews each; that is one region's process.
Folklore. That every Oliver Wyman case is a bank: the four practice cases on the firm's own preparation page are a grocer, a supermarket pharmacy, a dairy farm, and an oil and gas pricing problem. That you need a finance degree or the CFA: you need arithmetic and the timing insight above. That McKinsey, BCG, and Bain never give bank cases: their financial institutions practices do, and none publishes its case mix. Forum threads and review sites are where these stories live, not evidence.
Practice this today
Take the Cedar Valley numbers and change two inputs: year-two charge-offs of 3.5% instead of 4.8%, and servicing costs of 2.0% on the near-prime book. Compute the combined return on equity and the new break-even loss rate, then deliver the recommendation aloud in under 60 seconds, ending with the threshold and one risk. If the arithmetic is not clean inside three minutes, run this first:
Then run a full financial services case from the case library or the live voice case with Ned, and listen for the moment you name a lever before you have named the constraint. That is where the points leak. Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
CoachNed is independent and not affiliated with Oliver Wyman, McKinsey, BCG, Bain, or any other firm mentioned here.
Frequently asked questions
What makes a financial services case interview different from a normal profitability case?
The cost of what the client sold is unknown at the time of sale, and a regulator sets how much equity must sit behind every sale. Your structure needs a branch for loss or claims timing and one for funding and capital, or it will recommend growth that lowers return on equity.
Do I need to know banking regulation to pass a consulting case?
No. You need to know that minimum capital ratios exist, that risk weights are given rather than invented, and that capital has a cost you can turn into a hurdle margin in one line. Reciting Basel numbers earns nothing on the sheet.
How is a financial services case different from a fintech case?
A fintech case is a growth story measured in users, volume, and take rate, with the balance sheet belonging to a partner bank. An incumbent financial services case is a balance-sheet story measured in spread, losses, and capital. The fintech guide covers the first; this page covers the second.
Sources
- Basel Framework, RBC20: Calculation of minimum risk-based capital requirements — 4.5% CET1 minimum and 2.5% conservation buffer. Checked 2026-09-24.
- EIOPA Solvency II single rulebook, Article 101 — SCR calibrated to a 99.5% one-year value at risk. Checked 2026-09-24.
- Federal Reserve, Review of the Supervision and Regulation of Silicon Valley Bank, key takeaways — March 9, 2023 outflows of over $40 billion, $100 billion more expected on March 10, asset growth from $71 billion to over $211 billion, reliance on uninsured deposits. Checked 2026-09-24.
- FDIC press release, second quarter 2026 Quarterly Banking Profile — industry net interest margin of 3.32%, 4,238 institutions. Checked 2026-09-24.
- FDIC Quarterly Banking Profile, second quarter 2026, statement and charts — net interest margin by asset-size group. Checked 2026-09-24.
- Oliver Wyman, Interview preparation — the firm's description of its case interviews, six-step approach, and four published practice cases. Checked 2026-09-24.
- Oliver Wyman, Entry-level careers in Latin America — Brazil and Mexico process: 26-question, 50-minute test, 15-minute fit interview, two case rounds of two interviews. Checked 2026-09-24.