Mental Math for Investment Banking Interviews

Introduction

Investment banking interviews are one of the few professional settings left where you are expected to do arithmetic in front of another person with no calculator in the room. Phones stay in pockets, the interviewer says a number out loud, and you are supposed to come back with something sensible in a few seconds.

Candidates find this surprising, because the job itself is done in Excel. The reason it happens anyway is that the arithmetic is a proxy for something else: whether you carry a working feel for magnitudes. An analyst who cannot tell instantly that $420 million of EBITDA at 9.5 times is close to $4 billion of enterprise value will not notice the afternoon a model spits out $40 billion, and noticing that is a large part of the first two years.

This is a working playbook for the arithmetic that actually appears in banking and private equity interviews: percentage and rounding shortcuts, awkward-number tricks, the multiples math under every valuation question, margin and growth relationships, compounding and the rule of 72, the IRR rules of thumb that turn a MOIC and a hold period into a return, fast accretion checks, back-of-envelope interest, and estimation questions.

Every example shows the steps, because the technique is the content. The closing sections cover the part candidates neglect: how to say the math out loud, how to build speed in the weeks before a superday, and the traps that turn a correct method into a wrong answer.

Nobody in banking believes an associate should compute a levered free cash flow schedule by hand. The calculator is absent for a narrower reason. Deal work happens in conversation, on client calls, in a car on the way to a meeting, and in the ninety seconds after a managing director asks what the target would be worth at eleven times instead of nine. The people who are useful in those moments are the ones who can produce a defensible number immediately and flag it as approximate.

What the Arithmetic Actually Signals

An interviewer watching you multiply is reading three things at once, and only one of them is the answer.

The first is order-of-magnitude control. Answering $4.1 billion when the true figure is $3,990 million is a rounding difference nobody minds; answering $399 million is disqualifying, because it means the magnitude is not anchored anywhere in your head. The second is method. A candidate who says "nine and a half times four hundred and twenty is ten times four twenty, less half of four twenty" has shown a repeatable process, and the interviewer now knows the next question will also be answered. A candidate who stares at the ceiling and produces the right number has shown nothing transferable. The third is composure: whether pressure makes you slower and quieter or faster and sloppier.

Where It Shows Up, and How Trading Interviews Differ

In a banking or private equity process the arithmetic is embedded rather than isolated. It appears inside the paper LBO, where you have five to ten minutes to reach a MOIC and an IRR. It appears in valuation questions, usually phrased as a chain: here is EBITDA, here is a multiple, here is net debt, here is the share count, what is the price per share. It appears in brainteasers and market sizing questions, which are graded on structure rather than on the number. It appears across a superday of back-to-back interviews, where the fifth interviewer asks the same style of question as the first and fatigue is the real variable. And it appears in modeling tests, not as a formal question but as the sanity check that separates candidates who catch a broken link from those who hand in a model with a $40 billion typo in it, which is one of the things interviewers look for in Excel modeling tests.

Trading and market making firms test something genuinely different, and it is worth naming the contrast so you calibrate your preparation correctly. Those firms commonly run timed arithmetic screens with dozens of questions in a handful of minutes and a hard speed threshold, sometimes with a penalty for wrong answers. That is a raw computation test. Banking is not that. No banking interviewer is timing you to the second or scoring you against a percentile of arithmetic speed. The banking bar is lower on velocity and higher on judgment: correct method, sensible rounding, right magnitude, and a clear explanation of what you just did.

The single highest-return piece of memorization in a private equity or sponsors-coverage process is the relationship between a multiple of money and an annualized return. Every paper LBO ends here, and interviewers ask the reverse constantly ("if we need a 25 percent IRR over four years, what do we have to sell it for?").

The exact relationship for a single investment with one cash outflow and one cash inflow is:

where $n$ is the hold period in years. You cannot take a fifth root in your head, so you memorize the grid instead. The figures below are the true values, rounded to one decimal.

If you memorize nothing else, memorize the rounded anchors, because a rule of thumb you cannot recall instantly is useless: 2x in five years is about 15 percent, 2.5x in five years is about 20 percent, 3x in five years is about 25 percent, and 2x in three years is about 25 percent. The table shows the exact values so you can see how close the round numbers sit; in the room you use the round numbers and say they are approximate. Those four anchors cover most of what gets asked.

Interpolating Between the Anchors

Straight-line interpolation inside a column is accurate to within about two tenths of a point across the range that matters, which is far tighter than the precision of the assumptions feeding it.

Take a compact example. A sponsor buys a business at 9.0 times $250 million of EBITDA, so enterprise value is $2,250 million, funds it with 5.0 turns of debt ($1,250 million) and writes a $1,000 million equity check. EBITDA compounds at 6 percent for five years to about $335 million, the exit multiple is flat at 9.0 times, giving an exit enterprise value of $3,015 million, and $100 million a year of debt paydown leaves $750 million outstanding. Exit equity is $2,265 million, so the MOIC is 2.27 times.

Now interpolate. In the five-year column, 2x is about 15 percent and 2.5x is about 20 percent, a spread of 5 points across 0.5 turns of MOIC. You are 0.27 of the way up, roughly half the gap, so add a little under 3 points to 15 and answer about 18 percent. The exact figure is 17.8 percent, and nobody in the room will mind the difference. For the full mechanics of getting to that MOIC in the first place, our walkthrough of how to complete a paper LBO in under five minutes runs the whole sequence.

Most interview arithmetic is percentage arithmetic wearing a costume. Margins, growth rates, interest, accretion, ownership, and fee sizing are all the same operation. The candidates who are fast are not computing faster; they are decomposing the problem into pieces that require no computation at all.

Break Every Percentage Into Pieces

The core move is to build the percentage you need out of 10 percent, 5 percent, and 1 percent, which are all free.

That last example is the pattern that carries the most weight: round to a number you like, compute, then apply a correction in the opposite direction. The correction is almost always smaller and easier than the original problem, and doing it explicitly keeps you from losing track of which way your error runs. Say the direction out loud as you go ("I rounded the base up, so my answer is a touch high"), because it converts an approximation into a controlled approximation.

The Fraction and Reciprocal Anchors

A second layer of speed comes from knowing the fractions that appear over and over in finance, so that a percentage becomes a division you already know.

The second and third rows are not arbitrary. They are earnings yields, the reciprocals of the P/E multiples you will be quoted in accretion and dilution questions, and having them memorized turns a two-step problem into a one-step problem. A company at 13 times earnings has a 7.7 percent earnings yield, and you did not divide anything.

Interview numbers are chosen to be slightly hostile: 9.5 times, 7.8 percent, 285 million shares. Two techniques handle nearly all of them.

Move to a Friendly Neighbour, Then Correct

Any multiplication by a number near a round one becomes two easy steps.

Two algebraic shortcuts are worth having as well, because they look like magic and cost nothing. Any pair of numbers straddling a round midpoint collapses into a difference of squares: 48 x 52 is 50 squared minus 2 squared, or 2,496, and 97 x 103 is 10,000 minus 9, or 9,991. And any number ending in five squares instantly: take the leading digits, multiply by the next integer up, and append 25, so 85 squared is 8 times 9 followed by 25, or 7,225.

Divide by Anchoring on a Product You Know

Division is where candidates stall, because long division out loud is slow and visibly painful. The fix is to stop dividing and start hunting for the nearest product you already know, then convert the remainder into a decimal.

Suppose enterprise value is $2.7 billion and EBITDA is $325 million, and you are asked for the implied multiple. Do not divide. Ask what 325 times eight is: 2,600. The remainder is 100, and 100 divided by 325 is a bit under a third, so the answer is about 8.3 times. Check it in the other direction if you have a second: 325 times 8.3 is 2,697.5, which is the number you started with.

The same move handles share prices. Equity value of $6.3 billion across 285 million shares: 285 times 22 is 6,270, the remainder is 30, and 30 divided by 285 is roughly 0.1, so the price is about $22.10 per share. You never performed a division; you performed one multiplication and one small fraction.

This is the arithmetic that shows up most often, because it is the arithmetic of the job. Almost every technical question that involves a number involves one of three chains: from EBITDA to enterprise value, from enterprise value to a share price, or from revenue growth to EBITDA growth.

Enterprise Value, Implied Multiples, and Share Price

The three operations are the same equation read in different directions, and you should be able to run any of them cold.

Forward: enterprise value equals EBITDA times the multiple. $420 million at 9.5 times is 420 times ten less half of 420, which is $3,990 million, and you say "call it $4 billion".

Backward: the implied multiple is enterprise value divided by EBITDA, handled with the anchoring technique above. Interviewers like this direction because it forces you to know which metric pairs with which multiple, a point our overview of the valuation multiples used in practice covers in detail.

Down the stack: enterprise value less net debt is equity value, and equity value divided by diluted shares is the price per share. Reverse it and you have the standard "the stock is at this price, what multiple is the market paying" question.

Margins and Growth Rates in One Pass

Margin questions are recognition problems. If revenue is $1,240 million and EBITDA is $310 million, do not divide: notice that a quarter of 1,240 is 310, so the margin is exactly 25 percent. Train yourself to test the obvious fractions first (a half, a third, a quarter, a fifth, an eighth) before reaching for anything harder.

Growth rates yield to the 1 percent unit. Revenue goes from $4,150 million to $4,610 million, an increase of $460 million. One percent of the base is 41.5, and 460 divided by 41.5 is a bit over eleven, so growth is about 11.1 percent. Computing 1 percent of the base first turns every growth question into a small division by a number you chose.

Linking Revenue Growth to EBITDA Growth

The most useful relationship in this whole section is the one connecting top-line growth and margin expansion to EBITDA growth, because interviewers ask it constantly in operating-case form. The approximation is:

EBITDA growth is roughly revenue growth plus the margin change divided by the starting margin.

Work it through. Revenue grows 8 percent and the EBITDA margin expands from 25 percent to 26 percent. The margin lift is one point on a 25 point base, which is a 4 percent relative improvement, so EBITDA should grow by roughly 8 plus 4, or about 12 percent. Check it longhand: $1,240 million of revenue growing 8 percent is $1,339 million, at a 26 percent margin that is $348 million, against $310 million before, which is 12.3 percent growth. The shortcut runs a fraction low because it drops the cross term, and the gap is immaterial for an interview answer.

Compounding is where mental math stops being arithmetic and starts being finance, and it is the area where candidates most often reach for a linear answer that is badly wrong.

Where the Rule of 72 Comes From

Setting a compounded balance equal to double the original and solving for the number of periods gives:

The approximation in the second step uses the fact that the natural log of one plus a small rate is close to the rate itself. That produces a rule of 69.3, which is the right number for continuous compounding but an awkward one to divide by. Seventy-two is close enough for annual compounding, slightly more accurate in the 6 to 12 percent range where most finance questions live, and divisible by 2, 3, 4, 6, 8, 9, and 12, which is the entire reason it won. The University of Northern Iowa mathematics department's note on continuous compounding walks through the same derivation and the reason 69 is the better constant when compounding is continuous.

Rule of 72

A mental shortcut for compounding: divide 72 by an annual percentage growth rate to estimate the number of years an amount takes to double. At 9 percent the rule gives 8 years against a true 8.04 years, and at 6 percent it gives 12 years against a true 11.9 years. Related constants extend it: divide 114 by the rate for the years to triple, and 144 for the years to quadruple.

The rule has two well-behaved cousins that almost nobody quotes and interviewers notice immediately. Divide 114 by the rate for tripling time: at 10 percent that is 11.4 years against a true 11.5. Divide 144 for quadrupling: at 10 percent that is 14.4 years against a true 14.5. Both follow from the same derivation with the log of three and the log of four in the numerator.

Compound Growth Without a Calculator

For projecting EBITDA or revenue forward a handful of years, the second-order expansion is close enough and easy to run:

In practice that means: take the simple product of rate and years, then add a correction equal to the number of pairs of years times the rate squared. Five years at 5 percent gives 0.25 plus ten times 0.0025, so 1.275, against a true 1.2763. Five years at 10 percent gives 0.50 plus ten times 0.01, so 1.60, against a true 1.6105. Five years at 7 percent gives 0.35 plus 0.049, so 1.399, against 1.4026. The approximation always runs slightly low because it truncates the series, so nudge the answer up a hair and you will be inside a percent.

Faster still is memorizing the four five-year growth factors that account for most prompts: 5 percent compounds to 1.28, 6 percent to 1.34, 8 percent to 1.47, and 10 percent to 1.61. Those four numbers will carry you through most paper LBO exit EBITDA calculations without any computation at all.

Get the complete guide: Download our comprehensive 160-page PDF, access the IB Interview Guide covering valuation, LBO and merger frameworks with the worked arithmetic behind each one.

Interviewers rarely want a merger model in your head. They want a direction, a reason, and a rough magnitude, and there is a clean two-step route to all three. The mechanics behind it are laid out in our guide to accretion and dilution analysis; what follows is the version you can run without paper.

The Earnings Yield Test

Every acquisition swaps one stream of earnings for another. The question is simply whether what you buy yields more than what you give up.

Put numbers on it. The acquirer trades at 18 times earnings, so its earnings yield is 5.6 percent. It is buying a target at 14 times, an earnings yield of 7.1 percent. New debt costs 7 percent pre-tax, and at a 25 percent tax rate the after-tax cost is 5.25 percent. Both the stock-funded and the debt-funded version clear the bar, and the debt-funded version clears it by more.

Earnings Yield

Net income divided by market capitalization, which is the reciprocal of the P/E multiple. A company trading at 14 times earnings has an earnings yield of about 7.1 percent. Earnings yield is the fastest tool for accretion and dilution questions because it converts a multiple into a rate that can be compared directly against the after-tax cost of the debt or cash used to fund an acquisition.

Sizing the Accretion in Two Steps

The magnitude follows from the same spread. Multiply the yield gap by the purchase price to get incremental earnings, then divide by the acquirer's share count.

The target earns $210 million and is bought at 14 times, so the price is $2,940 million, funded entirely with 7 percent debt. The spread is 7.1 percent less 5.25 percent, call it 1.9 percent, and 1.9 percent of $2,940 million is roughly $56 million of incremental net income. The acquirer earns $900 million across 300 million shares, so standalone EPS is $3.00, and $56 million across 300 million shares adds about $0.19, which is roughly 6 percent accretion.

Verify it the long way and the two routes agree. Pro forma net income is 900 plus 210 less the after-tax interest of $154 million, so $956 million, and across an unchanged 300 million shares that is $3.19 per share, up 6.2 percent. The shortcut got there in two operations instead of six.

Leveraged finance and restructuring interviewers lean on interest math because it is the fastest way to see whether a candidate understands a capital structure or has just memorized the tranches.

Everything runs off 1 percent of the debt balance. On $850 million of debt at 8.5 percent, 1 percent is 8.5, so 8 percent is 68 and the extra half point is 4.25, giving about $72 million of annual interest. For floating-rate debt, add the base rate and the spread first: $1,200 million priced at a 4.25 percent base plus 375 basis points is 8.0 percent flat, so $96 million.

When debt amortizes during the year, use the average balance rather than the opening balance, which is the single most common source of overstated interest in a paper LBO. Debt falling from $850 million to $610 million averages $730 million, so interest at 8.5 percent is about $62 million, not $72 million. From there, coverage falls out immediately: $310 million of EBITDA against $72 million of interest is 4.3 times. And paydown compounds gently in your favor, because every $80 million of debt repaid saves roughly $7 million of annual interest at these rates, which becomes additional cash for repayment in the following year.

Estimation questions look like a different genre, but they are the same skill with the arithmetic made deliberately loose. The interviewer wants to watch you build a structure and populate it with defensible assumptions.

Market Sizing Question

An interview question that asks a candidate to estimate the total size of a market, a population, or a quantity with no data provided, using structured assumptions and round arithmetic. Also called a Fermi problem or an order-of-magnitude estimate, it is graded on the logic of the breakdown and the transparency of the assumptions rather than on the accuracy of the final number.

The method is divide and conquer: split the unknown into factors you can each estimate within a factor of two, then multiply. That decomposition is the same technique taught in engineering estimation courses such as MIT's Art of Approximation in Science and Engineering, where the discipline is choosing a breakdown whose errors partially cancel rather than compound.

A worked pass. Estimate the annual United States market for a consumer subscription priced at $15 per month. Start with population: the Census Bureau's population estimates put the country at roughly 340 million. Average household size is a little over 2.5 people, so households are about 136 million; call it 135 million, which happens to land within a percent of the published count of about 135 million, though you should say out loud that you would not have known that in the room. Assume 20 percent of households would subscribe, giving 27 million subscribers. Annual revenue per subscriber is $180, and 27 million times 180 is about $4.9 billion. State the number, then immediately name the assumption you are least confident in, which is the 20 percent penetration rate, and say what the answer becomes if it is half that.

The delivery is not decoration. In a live interview it is roughly half the grade, because it is the only window the interviewer has into whether you knew what you were doing or got lucky.

Announce the route before you run it. Saying "I will take ten times EBITDA and back off half a turn" costs two seconds and buys you a shared frame, so that if you fumble the subtraction the interviewer already knows the method was right. Then go quiet for the actual computation. Talking while multiplying is the single most reliable way to lose a digit, and interviewers are entirely comfortable with a four-second silence.

Say the rounding out loud, including its direction. "I rounded the share count down, so the price per share is slightly overstated" tells the interviewer you are tracking your own error, which is exactly the habit that makes an analyst trustworthy on a live model. Finally, deliver the answer as a number with a stated tolerance, and offer the refinement rather than performing it unasked.

If you catch an error partway through, name it and keep the framework. "I used entry EBITDA for the exit value, let me redo that one line" is a good look; silently changing a number is not, and abandoning the whole approach is worse. Interviewers see mistakes in every candidate they meet. What they are sorting on is whether the mistake is caught and repaired cleanly.

Mental math improves on a schedule that looks a lot like fitness: short sessions, most days, with progressive load. Four weeks of ten to fifteen minutes a day beats two marathon weekends, and the last week should be under a clock because timing changes how your brain behaves.

Interview arithmetic is a practiced skill, not a talent: Work through technical questions, brainteasers and estimation problems with full solutions, start practicing interview questions for free and find out which shortcuts you still reach for too slowly.

Almost every wrong answer in an interview comes from a small number of failure modes, and all of them are avoidable with a habit rather than more skill.

Nothing in this playbook is difficult mathematics. The percentage ladder, the friendly-neighbour correction, the earnings yield test, and the MOIC grid are all things a motivated candidate can hold comfortably in memory within a month. What makes them valuable is that they are retrieval rather than computation: in an interview you are not solving the problem, you are recognizing which of a dozen familiar shapes it belongs to and executing a move you have already run two hundred times.

That is why the practice sequence matters more than the list. Reading a shortcut and using a shortcut under mild social pressure with a stranger watching are different skills, and only one of them gets tested. Build the fluency in week one, layer the finance on in week two, memorize the grids in week three, and spend week four talking out loud with a timer running.

The payoff extends past the interview. The habits here (fixing units first, rounding deliberately, checking magnitude against a known anchor, saying what you assumed) are exactly the habits that keep an analyst from sending out a page with a decimal in the wrong place at two in the morning. Interviewers know that, which is the real reason the calculator never makes it into the room.