Financial Analysis interview preparation

Questions

100

Traced to a firm

42

Firms

28

Updated

September 2026

Asked at

All firmsMoody's7Bain Capital3SSState Street3AMAres Management2BLBlackRock2DED.E. Shaw2MSMorgan Stanley2Oaktree Capital Management2S&P Global2Bridgewater Associates1Citadel1FTFranklin Templeton1Golub Capital1HWHarris Williams1Houlihan Lokey1J.P. Morgan1Jane Street1MWMarshall Wace1Millennium Management1Morningstar1PIMCO1Sycamore Partners1TSTruist Securities1Two Sigma1Vanguard1WMWellington Management1Wells Fargo Securities1Wolverine Trading1

Topic

All topicsThree statements9Accounting policy and standards5Working capital and cash7Ratio analysis8Forecasting and budgeting9Variance and management reporting7Unit economics and costing8Capital budgeting7Cost of capital and valuation7Markets and rates5Modelling, Excel and data8Business partnering6Brainteasers and estimation4Fit and career10

Level

AnyCoreIntermediateHard

Type

AnyTechnicalCaseBrainteaserMarket viewFit

1. 031What does a complex revenue model look like, and how would you build one?Forecasting and budgetingHardtechnicalHoulihan LokeyInvestment Banking · New York · 2026

   Say this

   A complex revenue build is one where revenue emerges from several interacting drivers rather than a growth rate: cohorts, churn, pricing tiers, mix and capacity. You build it as a separate driver schedule feeding one revenue line, so the model stays auditable.

   Then walk it

   1. The structure I use: a driver tab with all assumptions hard-coded in one colour, a build tab that turns drivers into units and price, and a single revenue line that flows to the P&L. Nothing hard-coded in the build.
   2. For a subscription business the build is a cohort waterfall: opening customers, plus new adds, less churn, times average revenue per user, with ARPU differing by cohort and by plan. That gives you net revenue retention as an output rather than an assumption.
   3. For a capacity business it is capacity times utilisation times realisation, with a ramp schedule for new capacity. For retail it is store count times sales per square foot, with a maturity curve on new stores.
   4. The part that makes it complex rather than merely long is mix. Revenue can grow while realisation falls because the growth is in the cheaper tier. So I model mix explicitly and show blended realisation as an output.
   5. Then the controls. A checks row for every schedule, units reconciling to the previous period, revenue reconciling to the segment disclosure for history, and a one-page summary with growth decomposed into price, volume and mix.
   6. The discipline I would state: complexity has to earn its place. If adding a fourth driver does not change the answer by more than a percent or two, I take it out. A model nobody can explain in five minutes will not be used.

   Where candidates lose it

   Describing a big model rather than a structured one. The interviewer wants architecture: drivers separated from calculations, mix modelled explicitly, checks built in, and a justification for every layer of complexity. Volume of tabs is not sophistication.

   Expect next

   • How would you model churn for a cohort-based business?
   • How do you stop a model like that becoming unauditable?
   • Where would you hard-code and where would you formula-drive?

   Reported by candidates at Houlihan Lokey (Investment Banking, New York, 2026). Source: Wall Street Oasis.

2. 032How do you verify the validity of a client's pipeline to forecast revenue?Forecasting and budgetingHardcase studyHWHarris WilliamsInvestment Banking · Richmond · 2025

   Say this

   Test it against history rather than accepting the weightings. Take the pipeline as it stood 12 months ago, see what actually converted by stage, and apply those realised rates instead of management's. The gap between the two is your adjustment.

   Then walk it

   1. First, back-test. Pull the pipeline snapshot from four quarters ago, match it to closed business, and compute conversion by stage, by deal size and by sales rep. If stage-four deals converted at 45 percent while the model assumes 80, you have your answer.
   2. Second, check ageing. Deals that have sat in the same stage for three quarters are not pipeline, they are hope. I would strip or heavily discount anything past a normal cycle length.
   3. Third, look for hygiene problems: duplicate opportunities, deals with no close date or a close date that has been pushed four times, values entered as round numbers, and a bulge in the final quarter that mirrors the sales incentive calendar.
   4. Fourth, corroborate outside the CRM. Signed letters of intent, purchase orders, customer references, and for a diligence exercise, calls with two or three named prospects. Revenue that cannot be corroborated gets a haircut.
   5. Fifth, check coverage. Pipeline value over the target. Three times coverage on a 33 percent historical win rate is consistent; three times coverage on a 15 percent win rate is a miss waiting to happen.
   6. Then I would present it as a range: management case, back-tested case, and a downside using bottom-quartile conversion, with the bridge between them explained in one slide. The bridge is the deliverable, not the number.

   Where candidates lose it

   Accepting management's probability weightings and multiplying. Every CRM is optimistic near quarter end. The work is back-testing realised conversion by stage and stripping stale deals, and saying that is what gets you hired.

   Expect next

   • The sales head says your haircut is insulting. How do you handle it?
   • What if the CRM data only goes back two quarters?
   • How would this change for a business with three large customers?

   Reported by candidates at Harris Williams (Investment Banking, Richmond, 2025). Source: Wall Street Oasis.

3. 037Your forecast has missed by more than 10 percent three quarters running. What do you do?Forecasting and budgetingHardcase studyCorporate FP&ABusiness finance

   Say this

   Decompose the misses before changing anything. If the errors are all in one direction it is bias and the fix is process and incentives. If they are scattered it is variance and the fix is the model and the driver set. You cannot treat bias and variance the same way.

   Then walk it

   1. First, measure properly. Forecast error by line, by business unit, by owner, over eight quarters, with the sign preserved. Mean error tells you bias; mean absolute error tells you precision. Most organisations only track the second and then wonder why nothing improves.
   2. Second, separate the miss into volume, price, mix and timing. Three quarters of missing on timing is a completely different problem from missing on price, and the conversation goes to different people.
   3. Third, look at who submits the numbers and what happens to them when they are wrong. If sandbagging is rewarded and optimism is punished, you have designed the bias in, and no amount of model work will fix it.
   4. Fourth, fix the drivers. If revenue is forecast off a pipeline whose conversion assumption has never been back-tested, that is the error source. Replace judgement with realised rates wherever history exists.
   5. Fifth, change the output format. Move from a single number to a range with a named central case, and publish the forecast-versus-actual scorecard monthly with owners' names on it. Visibility corrects bias faster than any methodology change.
   6. The realistic expectation I would set: getting mean absolute error from 12 percent to 5 is a two- or three-quarter programme, not a month, and some businesses are genuinely unforecastable at that precision. Saying so is more credible than promising accuracy.

   Where candidates lose it

   Going straight to 'build a better model'. The most common cause is incentive-driven bias, not model error, and the diagnostic that separates them is whether the errors share a sign. Lead with that.

   Expect next

   • How would you present the bias finding to the business head who caused it?
   • What accuracy is realistic for a project business?
   • Would you change anyone's incentives?

Firm tags come from public, anonymous candidate reports on Wall Street Oasis: strong signal, not sworn testimony. Firms are named as the places a question was reported, not as partners of Fin Maverick. Answers are written for this page to show how to think out loud; they are not scripts to recite.