Financial Modeling Test: What to Expect and How to Practice
Updated: Nov 4
Mergers & Inquisitions notes that the median score on timed LBO modeling tests is often below 50%, which means a complete model with a few errors can still score well. A modeling test checks whether you can build a model that works and gives the right answer within the time limit, so preparing for one means building models from a blank sheet against a clock. You get a prompt and some financial data, and the firm asks for a working model in Excel, usually an LBO, a DCF, or a three statement model with a valuation on top.
Private equity firms give these tests most often, and banks and growth equity firms use versions of their own. By the end of this guide you'll know which of the five formats you're likely to face, what graders check first, how to budget your time so there's always an answer on the page, and how to practice in the weeks before, with a worked DCF example to build alongside.
Which Firms Give Modeling Tests, and When
Private equity firms give modeling tests most often, while banks use them less consistently and growth equity firms test a different kind of model. Mergers & Inquisitions says modeling tests and case studies come up in "virtually any private equity interview process," and that timed LBO tests of one to three hours are most common in on-cycle recruiting at large U.S. firms (M&I on the PE case study). They usually come late, after you've done well in earlier interviews, although many London funds start with a modeling test (M&I on PE interviews). Most PE candidates meet their first modeling tests while recruiting from an investment banking analyst seat, early in their analyst years, even though most don't start in PE until after about two years in banking.
Banks use modeling tests less consistently. M&I's three statement tutorial notes that banks like to test this model because it's a quick way to see who is proficient in Excel (M&I three statement model), and lateral candidates moving between banks have "a decent chance" of getting a modeling test or case study (M&I on lateral hiring). The type of test can also vary with the coverage or product group you're applying to.
Growth equity firms tend to test a different model. Their cases often ask you to forecast revenue from customer-level data, with renewals, upgrades, and downgrades, and the firms typically buy minority stakes rather than whole companies (M&I growth equity case study). Operationally focused PE firms and some emerging-markets funds also lean toward three statement tests (M&I on PE interviews).
The Five Test Formats Compared
Each format gives you different materials, a different clock, and a different idea of a good result. The format decides what you should practice, so find out which one you're getting before you start preparing. The recruiter or whoever scheduled the round can usually tell you whether it's on paper or in Excel and how long it runs.
Format | Who uses it | What you're given | Time | What's evaluated |
Paper LBO | PE firms and headhunters, as an early screen | A short prompt with a few round numbers; pen and paper, usually no calculator | About 5 to 30 minutes | Whether you understand how an LBO creates returns and can do the math by hand |
Timed LBO in Excel | PE, most of all in on-cycle recruiting at large U.S. firms | A prompt with the deal terms, sometimes an Excel template | 1 to 3 hours | Speed and accuracy: a finished model that answers the prompt's questions |
Timed three statement model, often with a DCF or LBO on top | Banks, lateral hires, operationally focused PE firms, growth equity, and PE firms in emerging markets | A packet of historicals, assumptions, and sometimes a CIM excerpt | 30 minutes to 4 hours | The right result and a model that's well structured |
Growth equity revenue case | Growth equity firms | Customer-level data rather than full historicals | About 1 to 2 hours | Whether you find the key drivers, test returns across cases, and judge if the plan is believable |
Take-home case and presentation | Smaller funds, off-cycle processes, and firms outside the U.S. | A company or a CIM and an open question about investing | 3 to 7 days | Your investment thesis, your presentation, and how you answer questions about both |
Sources for the table: Wall Street Prep on the paper LBO; M&I on the PE case study, PE interviews, the timed LBO test, the three statement model, and the growth equity case; and Wall Street Prep on the advanced LBO test.
Four of the five formats fit inside an afternoon, so most of your practice should be timed builds of one to three hours.
A paper LBO is closer to a long technical question than a model, so the practice is mental math and the LBO steps out loud; our walk me through an LBO post covers that answer. The timed tests can be on-site or sent over Zoom and email (M&I), so practice on a laptop setup like the one you'll use. The take-home is the one format where the model matters less than your thinking: M&I suggests spending more of your time on research and the presentation than on the model itself.
What's in the Case Pack
A timed test usually arrives as a prompt plus a packet of data. Wall Street Prep describes a longer PE test as a laptop, a prompt, and a 5 to 15 page packet of financial data you type into Excel yourself, often with a CIM (the confidential information memorandum a bank writes about a company for sale) and management projections (WSP). This is more likely at upper middle market firms and mega funds, in the later rounds of the process.
The prompt: what to build, the time limit, and the outputs the grader wants, such as an implied share price, an IRR, or a balance check. Write the requested outputs on paper before you type anything, because they decide what you build first.
The historicals: several years of historical financials, usually simplified to the lines the model needs.
The assumptions: the numbers you'll build from, such as the tax rate, debt terms, and exit multiple. Use the pack's numbers when it gives them; when it doesn't, make an assumption, type it in blue, and make sure you can defend each one later.
A CIM excerpt or template, sometimes: the CIM is the company's own story, with detail on its operations, financial performance, and growth drivers.
If the test uses a public company instead, the inputs come from its 10-K. The financial statements are in Item 8, "Financial Statements and Supplementary Data," and management's discussion of segments and results is in Item 7, "Management's Discussion and Analysis of Financial Condition and Results of Operations" (SEC Form 10-K). The debt note in Item 8 gives you each tranche's principal and rate.
How Graders Score a Modeling Test
Firms don't publish rubrics, but the prep sources agree on what matters. Graders look first at whether you finished and whether the answer is right; on longer tests they also judge how the model is built, and for take-home tests, they may ask for the rationale behind your assumptions. Wall Street Prep says that on its standard LBO test "the model structure is usually not part of the scoring," while on its three to four hour test scoring is "based on getting the right result, plus how well the model is structured" (WSP). M&I's advice for timed LBO tests is to finish the model, and it notes that the median score on these tests is often below 50% (M&I), so a complete model with a few errors can still score well.
Finished and working: the outputs the prompt asked for exist, the balance sheet balances or the sources equal the uses, and nothing shows an error. M&I's three statement tutorial says that if you can't complete everything, simplify so you can answer at least the main questions by the end (M&I).
Correct: a handful of cells decide whether a DCF gives the right answer, and a grader can check each one in seconds:
Working capital sign: an increase in net working capital is subtracted from free cash flow (WSP on unlevered FCF).
Tax line: tax is computed on EBIT rather than taken from the income statement (WSP on the interest tax shield).
Terminal value discount: the terminal value is a future number and has to be discounted back, using the full final year when you use an exit multiple (BIWS on the mid-year convention).
Discount periods: they match the convention you chose, such as 0.5, 1.5, and 2.5 for mid-year discounting (WSP).
The bridge to equity value: debt is subtracted and cash added, and the share price uses diluted shares.
In an LBO test, the equivalent cells are the ones Wall Street Prep's standard LBO test walks through:
Sources and uses: both sides balance once the sponsor's equity is filled in.
Debt paydown: any cash sweep pays down debt only from the excess free cash flow left after required payments.
Exit equity: the exit equity value deducts net debt at exit.
Returns: the IRR and multiple of money come from the sponsor's own cash flows.
Clean, on longer tests: blue font color for hard codes and black font color for formulas (WSP), one consistent formula across each row, units labeled, and no typed number hidden inside a formula. A grader who clicks a cell should be able to trace it back to an input.
Defended: WSP notes you may get questions about your opinion of the investment itself, and M&I says take-home cases are judged mostly on your thesis, your presentation, and how you answer questions afterward.
What to Know Before You Sit the Test
The test assumes you already understand the models and leaves the clock for building.
The mechanics of the model: how the three statements link, how unlevered free cash flow and WACC turn into a DCF, and how an LBO's debt paydown drives returns. Our three statement model for a modeling test, how an LBO model works, and walk me through a DCF posts cover each one in full.
How a model is laid out: inputs in one place, one formula per row filled across the years, check rows for every tie, and a toggle for any circular reference. Linking interest expense into the income statement creates a circularity, and WSP recommends an input cell that switches interest off and on so a single error doesn't spread across the model (WSP).
Excel speed: M&I says that without strong Excel skills you'll struggle to finish a 60-minute test (M&I). These shortcuts cover most of what a test asks your hands to do (Windows keys, from Wall Street Prep's shortcut list and Microsoft):
Shortcut | What it does | When you use it in a test |
Ctrl + arrow; Ctrl + Shift + arrow | Jump to the edge of a block; select to the edge | Grabbing a row of formulas before filling it |
Ctrl + R; Ctrl + D | Fill right; fill down | Writing a year's formula once and copying it across |
F2 | Edit the cell and color its references | Checking what a formula points at |
F4 | Cycle a reference through absolute and relative | Anchoring a rate or assumption cell |
Alt + = | Insert a SUM | Subtotals |
Ctrl + 1 | Open Format Cells | Number formats and font colors |
Ctrl + [ ; Ctrl + ] | Select a formula's precedents; its dependents | Tracing a number back to its input |
Ctrl + ` | Show formulas instead of values | Spotting typed numbers in formula rows |
Alt, A, W, T | Data table | The sensitivity grid |
Ctrl + Page Up / Page Down | Move between sheets |
Our Excel skills for a finance interview post goes deeper on functions and formatting.
A Worked Example: The Core of a Two-Hour DCF Test
The example below is the second half of a typical two-hour DCF test, from the operating projections to the share price, on a company made up for this article: Torvik Flow Systems, an industrial that makes pumps, valves, and controls. It's a DCF because that's a valuation banks and three statement tests often put on top, and the operating build and the checks carry over to the other formats; for a worked LBO, see how an LBO model works and WSP's standard LBO test linked above. The first half of the test (typing in historicals, projecting revenue, and building working capital and PP&E schedules) follows the same steps as our three statement model article, and our walk me through a DCF post explains the valuation logic behind the steps below, so here you start from the operating projections. All figures are in $ millions except per share amounts.
The prompt: Build a five-year DCF for Torvik as of December 31, 2025, with a WACC from the peer set provided, a terminal value by exit multiple with a perpetuity growth cross-check, an implied share price, and a sensitivity table of share price to WACC and exit multiple. You have two hours.
What the pack gives you: a 25% tax rate, a 4.50% risk-free rate, a 5.00% equity risk premium, a 9.5x exit multiple on final-year EBITDA, a $38.00 share price, 60.0 million diluted shares, debt of 400.0 in two tranches, cash of 95.0, and noncontrolling interest of 10.0. Your operating build has projected revenue of 1,426.6 in 2026E rising to 1,712.9 in 2030E, with growth slowing from 5.7% to 3.9%.
Step 1: From EBIT to Unlevered Free Cash Flow
Unlevered free cash flow is the cash the business produces before anyone who financed it gets paid. You start from EBIT, tax it as if the company had no debt, add back D&A because it reduced EBIT without costing cash, subtract capex because it costs cash that EBIT never counted, and subtract the increase in working capital because that's profit still sitting in receivables and inventory.
The tax line is on EBIT, the working capital line is negative every year, and the five present values add up to 627.9.
Take 2026E. EBIT is 211.1, and tax at 25% is 52.8, leaving NOPAT of 158.4. Add back 47.1 of D&A, subtract 57.1 of capex and the 13.6 increase in working capital, and unlevered free cash flow is 134.8. With the mid-year convention, the first year's cash is treated as arriving halfway through the year, so it's discounted 0.5 years: one divided by 1.09165 to the 0.5 power is 0.9571, and 134.8 times 0.9571 is a present value of 129.0. The later years use 1.5, 2.5, 3.5, and 4.5, and the five present values sum to 627.9.
The tax line is the one to watch. The income statement shows 43.0 of tax in 2025A because the company deducted 24.0 of interest first, but the DCF uses 49.0, the tax on EBIT. The benefit of deducting interest shows up in the discount rate through the after-tax cost of debt, so using the income statement's taxes would count it twice.
Step 2: The Discount Rate
WACC is the blended return Torvik's lenders and shareholders require, and in a test it's a few small calculations you can do in ten minutes once the pack's numbers are in.
Piece | How it's built | Result |
Pre-tax cost of debt | 250.0 at 5.75% and 150.0 at 6.50%, weighted by principal | 6.03% |
After-tax cost of debt | 6.03% times (1 minus 25%) | 4.52% |
Beta | Median unlevered beta of five peers (0.968), relevered at Torvik's debt to equity of 0.175 (400.0 over a 2,280.0 market cap) | 1.096 |
Cost of equity | 4.50% plus 1.096 times 5.00% | 9.98% |
Weights | Equity 2,280.0 and debt 400.0, out of 2,680.0 | 85.1% and 14.9% |
WACC | 85.1% times 9.98% plus 14.9% times 4.52% | 9.16% |
The model carries full precision, so recomputing from these rounded inputs can differ in the last digit (the beta comes out at 1.095, for example).
You unlever the peers' betas, take the median, and relever at Torvik's own debt because a raw beta mixes business risk with each company's debt load, and the median keeps one unusual peer from skewing the answer. You weight by market values because WACC is the return investors need on what their stakes are worth today.
Step 3: Terminal Value to Share Price
The terminal value stands in for every year after 2030. With the exit multiple method, you assume Torvik is sold at the end of 2030 for 9.5 times its 2030 EBITDA.
Of the 2,622.2 enterprise value, 1,994.3 comes from the terminal value, which is why graders push hardest on the exit multiple.
2030E EBITDA of 325.5 times 9.5 is a terminal value of 3,091.8. Because the sale happens at the end of 2030, it's discounted five full years even though the yearly cash flows use mid-year periods: one divided by 1.09165 to the fifth power is 0.6450, and 3,091.8 times 0.6450 is 1,994.3. Add the 627.9 from Step 1 and enterprise value is 2,622.2.
Enterprise value belongs to everyone who financed the business, so you move to the shareholders' slice by subtracting the 400.0 of debt, adding the 95.0 of cash, and subtracting the 10.0 of noncontrolling interest (the part of a subsidiary someone else owns). Equity value is 2,307.2, and divided by 60.0 million diluted shares that's $38.45 a share, about 1.2% above the $38.00 market price.
Step 4: The Cross-Checks and the Sensitivity Grid
Three quick checks tell you whether the answer is reasonable, and they're also the questions a grader is likely to ask.
Terminal value share: 1,994.3 of the 2,622.2 enterprise value is about 76%, so the exit multiple carries most of the valuation.
Implied growth: under the mid-year convention, a perpetuity growth terminal value is discounted half a year earlier than an exit multiple value, 4.5 years here, because the perpetuity is built on cash flows that arrive mid-year (BIWS). To find the growth rate the 9.5x multiple implies, restate its terminal value as of year 4.5 (3,091.8 divided by 1.09165 to the 0.5 power, about 2,959.2) and solve the growing perpetuity formula for growth: terminal value times WACC, minus the final-year cash flow of 178.1, divided by terminal value plus that cash flow. That gives about 3.0% a year forever after 2030. Running it the other way, 2.5% perpetual growth gives a terminal value of 2,739.7 (the final-year cash flow of 178.1 grown one year, divided by WACC minus 2.5%), about 8.4x EBITDA, and a share price of about $36.00. The two methods bracket the answer between roughly $36.00 and $38.45.
Market multiple: the market values Torvik at 10.8x its 2025A EBITDA today (2,595.0 of enterprise value over 240.0), so a 9.5x exit assumes the multiple shrinks as growth slows, which is the cautious side.
The sensitivity grid shows the share price as WACC and the exit multiple move. In Excel it's a two-variable data table (Data, What-If Analysis, Data Table, with a row input cell and a column input cell) (Microsoft).
The center cell matches the DCF at $38.45, and each half turn of multiple moves the price about twice as far as each half point of WACC.
Confirm the center cell first, because if it doesn't match the DCF, the input cells point at the wrong place. Then read it: each half turn of exit multiple moves the price about $1.75, and each half point of WACC moves it about 85 to 90 cents, so the exit multiple is the assumption to defend most carefully.
How to Budget Your Time Under the Clock
Build in the order that leaves an answer on the page at every point, so a model you don't finish still shows a share price. A grader can give credit to a DCF with no sensitivity grid, but a beautiful schedule with no valuation answers none of the prompt's questions. Here's one way to split a two-hour DCF test:
Time | What to build | Why it goes here |
0:00 to 0:08 | Read the prompt twice; write every requested output on paper | Building the wrong thing wastes the whole test, and a few minutes here prevent it |
0:08 to 0:25 | Inputs: historicals in blue, ratios, the assumptions block | Everything downstream points here |
0:25 to 0:45 | Operating build: revenue, costs, EBIT, D&A, capex for each projected year | The longest block; if you're running late, project one revenue line instead of segments |
0:45 to 0:58 | Working capital and PP&E schedules | They feed the cash flow |
0:58 to 1:10 | Cost of debt, beta, WACC | Mechanical once the pack's numbers are in |
1:10 to 1:30 | DCF through the share price | The answer the prompt asks for |
1:30 to 1:38 | Checks: terminal value share, implied growth, signs on working capital and net debt | The grader's first questions, answered by you first |
1:38 to 1:48 | Sensitivity grid, then confirm its center cell | The output that makes the debrief easier |
1:48 to 2:00 | Formatting pass, then save with the requested file name and reopen it | A clean file that's actually saved |
Save every fifteen minutes, and once more before you run the data table, because a data table is the slowest thing in the file to calculate. For a shorter test, cut detail from the operating build first; for a three-hour test, the extra time usually goes to a full balance sheet and cash flow statement under the valuation.
A timed LBO follows the same logic in its own order. Wall Street Prep's standard LBO test builds the assumptions and sources and uses first, then the income statement and cash flow, then the debt schedule, then returns and a sensitivity, because the returns depend on how much debt is left at exit. M&I's advice is to skip a full three statement model unless the prompt asks for one (M&I).
When time is tight, add the things a grader checks and skip the ones a grader merely notices:
Worth adding: check rows for every tie, the perpetuity growth cross-check, a one-row sensitivity if the full grid won't fit, and a short outputs block at the top with the share price.
Worth skipping unless the prompt asks: extra segment detail, a separate tax schedule, a full comps tab, quarterly periods, charts, and conditional formatting. M&I's timed LBO tutorial makes the same point about formatting: keep it bare-bones or you won't finish on time (M&I).
Common Modeling Test Mistakes and What They Do to the Answer
An undiscounted terminal value moves the share price by about half, while a flipped working capital sign or tax taken from the income statement moves it only a few percent, so those smaller errors are found only by checking the rows themselves.
Leaving the terminal value undiscounted turns a fairly valued stock into one with 49% upside, which is the first thing a grader would question.
Working capital added instead of subtracted: the price rises to $40.22, because the model now rewards the company for tying up more cash.
Terminal value left undiscounted: the 3,091.8 is added at face value and the price jumps to $56.74.
Tax taken from the income statement: with interest held at 24.0, this raises each year's cash flow by 6.0 (24.0 times 25%) and the price to $38.86, the hardest of the three to spot because the number looks reasonable.
A hardcoded plug: typing a number to force a balance check to zero hides a problem a grader can find by clicking the check row. A visible check that's off, with a one-line note on what you think is wrong, shows judgment instead.
Mismatched units or years: revenue in thousands next to debt in millions, or a September fiscal year treated as December. Write the units in each block's header and the fiscal year end at the top before you start.
A formula range that stops short: a SUM that misses a row inserted after it was written. Press F2 on each subtotal before you hand in to see what it covers.
How to Practice in the Weeks Before
The test rewards having built the same kind of model from a blank sheet enough times that the order is automatic, and M&I says proficiency with the key shortcuts and practice with many modeling tests are essential (M&I). Spend the first weeks learning one model's layout from a finished file, the next weeks rebuilding it under a clock, and the last week rehearsing the exact format the firm told you about.
Four weeks out, learn the layout from a finished model. Start with a worked case that shows the prompt, the inputs, and the finished file, such as M&I's 90-minute three statement case on Otis, which comes with the prompt, the company documents, and a completed Excel file, or Wall Street Prep's standard LBO modeling test, which walks through a prompt and a step-by-step solution (its Excel template sits behind a sign-up form). Trace every formula with Ctrl + [ until you can say what each block does, then close the file and rebuild it from a blank sheet, untimed. This stage builds the map of the model: you need to understand what each part does and how each input flows through to the valuation, so speed doesn't matter yet.
Two to three weeks out, rebuild under the clock. Build the same model from a blank workbook, timed, with no notes, and write down where you slowed down. The second and third builds show you which blocks cost you the most time, and those are the ones to drill between builds. Then try one build on a real company: pull a 10-K from the SEC's EDGAR database (or from S&P Capital IQ or FactSet if your school has access), and pick a stable company, ideally one with a simple capital structure and a December year end, so you practice finding the inputs as well as modeling them.
The last week, rehearse the real format. Do a few timed builds in the exact format and length the firm told you about, ideally at the time of day the test is scheduled, and practice without the mouse so every move is a keystroke. Before you hand in each practice model, run the audit a grader would: show formulas with Ctrl + ` and look for typed numbers in formula rows, trace the share price back to its inputs, set the terminal growth or multiple to an extreme and confirm the price moves the right way, and check the units and valuation date against the prompt. Then have a friend open your file and ask you why your exit multiple and WACC are reasonable.
Questions to Expect After You Hand It In
Is 9.5x a reasonable exit multiple? The pack set it, and it's cautious next to the market, which pays 10.8x today, and revenue growth slows from 5.7% to 3.9% over the projection, so a buyer in 2030 should pay somewhat less. It's also the assumption that matters most, since about 76% of enterprise value is terminal value and each half turn is worth about $1.75 a share.
Does the multiple make sense against long-run growth? It implies about 3.0% growth a year forever. If you think 2.5% is the right long-run rate, the perpetuity method gives about $36.00, so next to that, the multiple is slightly generous, and the honest range is roughly $36 to $38.50.
Why mid-year discounting? Torvik's cash comes in through the year, so treating it as arriving at the midpoint is closer to reality; here it adds about 45 cents a share compared with end-of-year discounting.
Your check doesn't balance. What happened? Say how far off it is, whether the gap is constant or growing (a constant gap usually means a missing link, and a growing one usually means a sign error), and what you'd fix with ten more minutes.
Comments