What Excel Skills You Need for a Finance Interview (2026)
How much Excel a finance interview tests depends on the seat, but the skills underneath stay the same. The Excel skills you need for a finance interview are fast keyboard navigation, a short list of formulas, and the core models (a three-statement model, a DCF, and an LBO) built well enough that you understand every line.
I'm Stephen Turban, founder of WSG. I went to Harvard and worked at McKinsey, and I've run a lot of mock technical interviews with students headed into banking.
If you want structured practice before a modeling test without paying for a course, start with these free investment banking courses here. FP&A interviews lean on Excel more directly than most finance roles, and the questions they ask are covered here.
How Do Finance Interviews Test Excel?
At an undergraduate banking superday, expect to explain a model out loud; in private equity, expect a paper LBO, a timed LBO build, or a take-home case.
Banking: a modeling test is unusual at an undergraduate superday. Mergers & Inquisitions' superday guide says you shouldn't expect a three-statement modeling test or an LBO model unless you're a lateral hire. The technical questions are asked out loud instead, and some of them ask you to walk through a model: the University of Maryland's 2026 investment banking interview guide lists "Walk me through an LBO model" among its sample questions. That's why building the models still pays off for a banking candidate, since a line you've typed yourself is much easier to explain.
Private equity: firms tend to test the build directly. M&I's private equity recruiting guide notes that some PE firms now hire undergraduates as analysts, and that PE interviews generally include case studies and modeling tests, which come in three formats:
Format | What you do | Time | Where it's most common |
Estimate the IRR and money multiple of a simple buyout with pen and paper | 5 to 10 minutes | Early rounds | |
Build an LBO in Excel, from a template or a blank sheet | 1 to 3 hours, on-site or over Zoom and email | On-cycle recruiting (the fast, early process for banking analysts) at large US funds | |
Research a company, build a model, and recommend for or against buying it | A few days up to a week | Smaller funds, off-cycle recruiting, and outside the US |
None of these guides says which format an undergraduate analyst role uses, so ask the recruiter, and prepare for both the paper LBO and the timed build.
The Keyboard Shortcuts to Learn First
Learn the Windows shortcuts for jumping and selecting (Ctrl and Shift with the arrow keys), editing (F2), anchoring references (F4), and pasting values (Ctrl + Alt + V) first, because they cover most of the moving, selecting, and editing you do in a model. Speed is why they come before formulas: M&I's case study guide calls the timed LBO test "a pure speed test" and says proficiency in the key Excel shortcuts is essential. Even without a timed test, they make the models below much faster to build. Practice on a Windows machine if you can get access to one, even if you own a Mac, since Wall Street Prep's shortcut cheat sheet points out that most of the Alt-key ribbon sequences have no Mac equivalent.
The full set of twelve adds sheet navigation, formatting, and two ribbon sequences:
Ctrl + arrow key: jump to the edge of the data in that direction
Ctrl + Shift + arrow key: select everything from here to the edge
F2: edit the active cell and see its formula
F4: cycle a reference through $A$1, A$1, $A1, and A1
Ctrl + 1: open Format Cells
Ctrl + Alt + V: open Paste Special, where you can paste values only
Ctrl + Page Up and Ctrl + Page Down: move between sheets
Alt + =: insert a SUM formula
Ctrl + [: jump to the cells a formula pulls from
Ctrl + Space and Shift + Space: select the whole column or row
Freeze panes: press Alt, W, F, F, one key at a time
Data table: press Alt, A, W, T, one key at a time
Microsoft keeps the full list on its keyboard shortcuts page.
The Formulas That Run Most Models
Beyond arithmetic and SUM, five groups of functions do most of the work in a model: MIN and MAX, IF, INDEX and MATCH (or XLOOKUP), IFERROR and ROUND, and XIRR.
MIN and MAX: a cash sweep, where spare cash prepays a loan, is the MIN of the cash available and the loan balance. With $60 million of spare cash and a $50 million balance, MIN(60, 50) repays $50 million, so the balance can't go below zero.
IF: one condition, one answer each way. A revolving credit line is drawn only when cash would go negative, so the draw is IF(cash before revolver < 0, -cash before revolver, 0).
INDEX and MATCH, or XLOOKUP: pull a value from a table by its labels. INDEX(B2:D4, MATCH("South", A2:A4, 0), MATCH(2026, B1:D1, 0)) returns South's 2026 revenue, with one MATCH finding the row and the other the column. VLOOKUP uses a fixed column number instead, so inserting a column can break it. XLOOKUP does the same job more simply in Excel 2021 or later and Microsoft 365.
IFERROR and ROUND: use both only on rows you present. On a calculation row, IFERROR would hide a broken formula, and rounding would make totals stop matching.
XIRR: calculates a return using actual dates, which matters when the periods aren't even. It needs the investment as a negative number and at least one positive cash flow after it.
How to Set Up a Model So It Changes Cleanly
Put every assumption in its own labeled cell and point every formula at that cell, so you can change an assumption once and watch it flow through.
That only works across copied formulas if the references are anchored with dollar signs. Say this year's revenue sits in B5 and the growth rate in B3, so =B5*(1+B3) in C5 grows revenue by that rate. Copy it one column to the right and it becomes =C5*(1+C3), which points at an empty cell, so growth silently drops to zero. Written as =B5*(1+$B$3), every copy reads the same growth rate. A typed number fails in a different way: =B5*1.05 copies fine, but the cell stops responding when you change the growth assumption.
Color coding lets anyone reviewing your model see which cells are inputs. House styles vary, but the common convention is blue text for hardcoded inputs, black for formulas, and green for links to another sheet. Put the unit in every header, like Revenue ($mm), so the scale is clear at a glance.
Which Excel Skills Can You Skip?
For a timed build or a model walk-through, you can usually skip pivot tables, macros, VBA, Power Query, and any chart beyond a simple bar or line. None of them appear in a three-statement, DCF, or LBO build. A take-home case can hand you raw data, though, so learn basic pivot tables if one is on your calendar.
Build a Three-Statement Model That Balances
Build the three-statement model first, because the DCF is built from it, and its balance check is a quick way to find a broken link.
The model balances because every cash flow line is the cash side of a change somewhere on the balance sheet. Net income, less any dividends, is the increase in retained earnings. Depreciation and capex together explain the change in property and equipment. An asset that grows uses cash, so a rise in working capital shows up as a negative number. Repaying debt uses cash and lowers debt by the same amount. Add those lines up and you get the change in cash, so both sides of the balance sheet move by the same amount.
Here's one year for a made-up company, in millions.
Opening balance sheet: cash $50.0, working capital $50.0, and property and equipment $150.0, against debt $200.0, paid-in capital $30.0, and retained earnings $20.0.
Assumptions: last year's revenue $500.0, growing 6 percent; EBITDA margin 20 percent; depreciation and capex $15.0 each; working capital 10 percent of revenue; debt at 8 percent, with $20.0 repaid; tax rate 25 percent; no dividend.
Income statement:
Revenue: $500.0 × 1.06 = $530.0
EBITDA: $530.0 × 20 percent = $106.0
EBIT: $106.0 minus $15.0 of depreciation = $91.0
Pre-tax income: $91.0 minus $16.0 of interest (8 percent of the $200.0 opening debt) = $75.0
Net income: $75.0 minus $18.75 of tax = $56.25
Cash flow statement:
Cash from operations: $56.25 + $15.0 of depreciation minus $3.0 of working capital build (10 percent of the $30.0 revenue increase) = $68.25
Change in cash: $68.25 minus $15.0 of capex minus $20.0 of debt repaid = $33.25
Balance sheet:
Cash: $50.0 + $33.25 change in cash = $83.25
Working capital: 10 percent of $530.0 = $53.0 (up $3.0, the build on the cash flow statement)
Property and equipment: $150.0 + $15.0 of capex minus $15.0 of depreciation = $150.0
Debt: $200.0 minus $20.0 repaid = $180.0
Retained earnings: $20.0 + $56.25 of net income = $76.25
Assets: $83.25 + $53.0 + $150.0 = $286.25
Liabilities and equity: $180.0 + $30.0 of paid-in capital + $76.25 = $286.25
Balance check: 0
Build the check under the balance sheet first, then add each cash flow line together with the balance sheet line it changes, so a break shows up at the link you just added.
A zero only shows your links are consistent. Forget to subtract depreciation from property and equipment and the check reads $15.0, but type a 30 percent tax rate by mistake and it still balances, at net income of $52.50, because the wrong number flows through every link the way a right one would. So check your inputs against the case materials too.
Build a DCF and a Sensitivity Table
A DCF values a company from the unlevered free cash flow your three-statement model produces, and its sensitivity table shows how far WACC and growth move the share price. It's one of the models a banking interviewer is most likely to ask about: M&I's interview questions guide names accounting, equity value and enterprise value, and valuation and DCF as the most common topics in entry-level interviews.
Where the cash flow comes from: the DCF starts from unlevered free cash flow, which you build from the three-statement model: EBIT after tax, plus depreciation, minus capex, minus the increase in working capital. Interest stays out because this cash belongs to lenders and shareholders together, so tax is taken on EBIT, as if the business had no debt. That's also why it's discounted at the weighted average cost of capital (WACC), and why net debt comes off at the end to get to the value for shareholders. In the three-statement example, that's $91.0 × 75 percent = $68.25, plus $15.0, minus $15.0, minus $3.0, for $65.25 million.
The example: a separate company with simpler numbers, in millions, valued at a 9 percent WACC and 2.5 percent long-term growth.
Free cash flow: $100.0 next year, growing 5 percent a year: $100.0, $105.0, $110.3, $115.8, and $121.6
Present value of years one to five, dividing each year's cash flow by 1.09 raised to the number of that year: $426.3
Terminal value, which captures everything after year five: year five's unrounded $121.55 × 1.025 ÷ (9 percent minus 2.5 percent) = $1,916.8. Multiplying by 1.025 gives year six's cash flow, where the growing perpetuity starts, and the result is a value at the end of year five.
Present value of the terminal value, discounted five years at 9 percent: $1,245.8
Enterprise value: $426.3 + $1,245.8 = $1,672.0
Share price: ($1,672.0 minus $200.0 of net debt) ÷ 50 million shares = $29.44
Building the sensitivity table: use Data, What-If Analysis, Data Table. Put a link to the share price in the top-left corner, growth rates across the top, and WACCs down the side, then point Excel's row input cell at the growth assumption and its column input cell at WACC. Training The Street notes that the table has to sit on the same sheet as the inputs it changes, so keep those two assumptions on the valuation sheet.
WACC | 2.0% growth | 2.5% growth | 3.0% growth |
8% | $32.89 | $35.59 | $38.84 |
9% | $27.55 | $29.44 | $31.65 |
10% | $23.55 | $24.93 | $26.51 |
What moves the price: the terminal value is about three-quarters of enterprise value here, so whatever moves it moves the price. It divides by WACC minus growth, a gap of only 6.5 points, so a one-point move in either input changes the terminal value by 13 to 19 percent, more when the gap narrows than when it widens. WACC moves the price more, because it also discounts every year's cash flow: a point lower adds $6.15 to the price, while a point more growth, to 3.5 percent and just past the table's range, adds $4.82. The table's columns step only half a point, so they look closer together than the rows.
Build a Simple LBO
For private equity, practice a simple LBO until the debt schedule feels automatic, because it's the model a timed test asks you to build. It has four blocks, the same ones M&I's LBO test tutorial builds: the purchase price with sources and uses, an operating build down to cash flow, a debt schedule, and the returns. Don't build a full three-statement model unless the test asks for one, since the cash flow projections already give the debt schedule what it needs.
The deal (in millions): a company with $100.0 of EBITDA is bought at 8.0x, or $800.0. That's paid for with $500.0 of debt at 8 percent and $300.0 of equity. EBITDA grows 5 percent a year, depreciation and capex are $20.0 each, working capital doesn't change, and the tax rate is 25 percent. Every dollar of cash left over repays debt through a MIN sweep. Because depreciation equals capex and working capital is flat, the cash left for repayment equals net income.
Why interest uses the opening balance: interest on the average balance would make interest, net income, and the repayment depend on each other, which Excel flags as a circular reference. Charging interest on each year's opening debt avoids the loop, and you don't want to debug one against a clock.
Year | EBITDA | Interest | Cash that repays debt | Ending debt |
1 | $105.0 | $40.0 | $33.8 | $466.3 |
2 | $110.3 | $37.3 | $39.7 | $426.5 |
3 | $115.8 | $34.1 | $46.2 | $380.3 |
4 | $121.6 | $30.4 | $53.3 | $327.0 |
5 | $127.6 | $26.2 | $61.1 | $265.9 |
Rows are rounded, so a total can differ by $0.1.
Returns and what drives them: at exit, the buyer sells at the same 8.0x on year-five EBITDA of $127.6, for $1,021.0 of enterprise value. Repaying the remaining $265.9 of debt leaves $755.2 of equity. That's a money multiple (MOIC) of 2.52x on the $300.0 invested, and an IRR of 20.3 percent over five years: =(755.2/300)^(1/5)-1, or XIRR on the two dated cash flows.
Of the roughly $455 gain, about $234 comes from paying down debt ($500.0 minus $265.9) and $221 from EBITDA growth at the same multiple (($127.6 minus $100.0) × 8.0), and that split is your answer when an interviewer asks what drives the return.
Put a data table on the returns too: at an 8.0x entry, exiting at 7.0x instead of 8.0x cuts the IRR to 15.9 percent, and exiting at 9.0x lifts it to 24.1 percent.
The same deal as a paper LBO: this is how you'd estimate it in your head, and the four steps work on numbers you haven't seen.
Exit value: grow EBITDA a year at a time in round numbers ($105, $110, $116, $122, and $128), which takes seconds and works at any growth rate. At 8x, the exit is about $1,025.
Year-one cash: ($105 minus $20 minus $40 of interest) × 75 percent, or about $34. That's net income, which equals the cash that repays debt here because of how the deal is set up; when capex differs from depreciation or working capital changes, add back depreciation and subtract capex and any working capital increase.
Year-five cash: estimate year-five debt as the opening $500 minus four years of year-one cash, or about $365, so interest is about $29 and cash is ($128 minus $20 minus $29) × 75 percent, or about $59. The estimate barely matters: leaving interest at $40 still gives about 2.5x.
Equity and IRR: cash grows by roughly the same amount each year, so the average of the first and last years is close to the true average. Average the two (about $46.5) and multiply by five, so about $233 of debt is repaid and about $267 is left. Equity is about $758, or about 2.5x your $300. Breaking Into Wall Street gives doubling your money in five years as about 15 percent and tripling it as about 25 percent, so 2.5x lands around 20 percent, close to the 20.3 percent the model gives.
How to Spend a Two-Hour Modeling Test
If a firm gives you a timed LBO test, build the blocks in the order cash flows through them, keep every driver as simple as the packet allows, and leave the last 20 minutes for checking. You don't need a perfect model: M&I's LBO test tutorial says you don't need a 100 percent score, and that finishing with most of it right puts you in good shape. For a one- or three-hour test, scale each block in proportion. In practice models, color inputs as you go, but under a clock keep formatting to number formats, as the same tutorial advises.
Minutes 0 to 10: read the whole packet and list every output it asks for. If an input is missing, put your assumption in a labeled cell.
Minutes 10 to 20: lay out the blocks with headers (assumptions, sources and uses, operating build, debt schedule, and returns), or learn where the inputs live if you're given a template.
Minutes 20 to 50: build sources and uses and the operating build down to cash available for debt repayment, since everything after it reads from this block.
Minutes 50 to 80: build the debt schedule, with the sweep as a MIN against the remaining balance.
Minutes 80 to 100: build the exit, MOIC, and IRR, then one data table on the returns.
Minutes 100 to 120: check signs and units, and note your assumptions. If the IRR doesn't fit the multiple (about 15 percent for 2x and 25 percent for 3x over five years), look for a sign error or a broken link.
If the test is live, say what you're doing as you build, so the interviewer can follow your logic even if the clock runs out.
How to Walk an Interviewer Through Your Model
Walk an interviewer through any model in the same order: the structure, then the drivers, then the output and what moves it. That order lets the interviewer pick a block to dig into, and naming the drivers shows the numbers came from operating logic. For the LBO above, it might sound like this:
"I built a five-year LBO with four blocks: the purchase and its funding, the operating build, the debt schedule, and the returns. The company is bought at 8x EBITDA with $500 million of debt, five times EBITDA. EBITDA grows 5 percent a year, and all spare cash sweeps the debt, so debt falls from $500 million to about $266 million. At an 8x exit that's a 2.5x multiple and about a 20 percent IRR, and a little over half the gain comes from debt paydown. The exit multiple is a big lever: a 7x exit drops the IRR to about 16 percent."
Be ready for a follow-up on any of those claims. If you're asked why you exit at 8x, say that exiting at the entry multiple keeps the return from depending on the multiple rising, which is the conservative default.
A DCF follows the same order: free cash flow, terminal value, and discounting first, then growth and WACC as the drivers, then the share price and how far the sensitivity table moves it. Our answers to walk me through an LBO and walk me through a DCF cover the verbal versions in more depth.
A Two-Week Plan to Practice
Spend the first nine days on Excel tools and the three-statement model, then spend the last five mostly on the DCF if you're recruiting for banking or on the LBO if you're recruiting for private equity. Each stage uses what the one before it built, so keep the order.
Days 1 to 3, shortcuts (about 20 minutes a day): unplug the mouse and move around any spreadsheet with Ctrl, Shift, and the arrow keys, then add F2, F4, Ctrl + 1, and Paste Special. Everything later goes faster once the keys are automatic.
Days 4 to 6, formulas and setup (about 45 minutes a day): write one formula with each function on the list, then build a small segment table you pull from with INDEX and MATCH and with XLOOKUP. Set it up with blue inputs and anchored references, since every later model uses both.
Days 7 to 9, the three statements (about 90 minutes a day): build the model from the three-statement section for five years, holding the year-one assumptions constant, on an inputs tab and a model tab. Year-one net income should read $56.25 million and the balance check should read zero every year.
Days 10 to 14 if you're recruiting for banking (about two hours a day):
Day 10: build the separate DCF example with its data table and check that it reads $29.44, with corners of $23.55 and $38.84, so you know the discounting and the table are wired correctly.
Day 11: replace its free cash flow with links to your three-statement model's unlevered free cash flow, keep the 9 percent WACC and 2.5 percent growth, and use the opening balance sheet's debt minus cash ($200.0 minus $50.0, or $150.0) as net debt, with the same 50 million shares. The price will change, and now every number traces back to revenue.
Day 12: build the LBO from the LBO section untimed.
Days 13 and 14: rebuild the three statements and the DCF from a blank workbook, then explain each model out loud in under two minutes, the way an interviewer would hear it.
Days 10 to 14 if you're recruiting for private equity (about two and a half hours a day):
Days 10 and 11: build the LBO from the LBO section untimed and check it against 2.52x and 20.3 percent, then work the same deal in your head as a paper LBO.
Days 12 to 14: change the entry multiple, the debt, and the growth rate each day so you can't rely on remembered answers. Start with a five-minute paper LBO on the new numbers, then do a timed build from a blank sheet and compare the two answers. Shorten the clock as you go, from two hours to 90 minutes to 60, since some tests give you only an hour.
Build a DCF for private equity only if a firm's case asks for one.
If private equity is the goal, our private equity internship guide covers which firms hire undergraduates and when they recruit.
Comments