# Mental math for interviews: what to practise and how Canonical: https://romeoapps.app/en/numvolt/mental-math-for-interviews/ Author: Roméo Gambino · Published: 2026-09-10 · Updated: 2026-09-10 · App: Numvolt (https://romeoapps.app/en/numvolt/) How to prepare for interview arithmetic in consulting, finance and trading: estimation, zeros, percentages, growth, the rule of 72 and calculating out loud. ## What interview arithmetic actually tests Case interviews, finance assessments and trading tests use arithmetic in different ways, but they rarely test obscure tricks. What they tend to reveal is whether you can keep track of magnitude, move between percentages and absolute numbers, and stay accurate while someone watches. Formats vary by employer and change over time, so prepare for the skills rather than for a rumoured list of questions. Read what the employer itself publishes about its process, and treat everything else as anecdote. ## Pressure is part of the task Calculating in front of an interviewer is harder than calculating alone, and research suggests why. Beilock and Carr found that performance pressure harmed mathematical problem solving by consuming working memory, and that the harm was concentrated among people with high working memory capacity and on the problems that demanded it most. Ashcraft and Kirk found that people with high math anxiety showed smaller working memory spans during computation. The practical conclusion is to practise under some pressure before the day, and to prefer methods that hold fewer numbers at once. ## Manage the zeros separately Large numbers become manageable when you split them into a small product and a count of zeros. For 6,000 × 700, multiply 6 × 7 = 42, then count the zeros, three plus two, which gives 4,200,000. For 3.2 million × 45, compute 3.2 × 45 = 144 and keep the unit: 144 million. Treating every large number as a small number followed by a known number of zeros is the habit that prevents the most interview errors. ## Say the units every time Is it 144 million or 144 thousand? Revenue per month or per year? Keeping units attached to every intermediate result costs nothing and catches the errors that matter most, because a wrong digit gives an answer that is slightly off while a wrong unit gives one that is off by a factor of a thousand. When dividing, say what the result means: 45 million a year divided by 12 is 3.75 million a month. ## Estimate first, then refine Before any exact calculation, give a rough answer. 5,960 ÷ 31 is roughly 6,000 ÷ 30, which is 200. The exact answer is 192.3 to one decimal, and knowing it should be near 200 protects you from a result like 19.2. In many case discussions the rounded figure is all that is needed. Say that you are rounding, say in which direction, and move on. ## Build a market size from stated assumptions Sizing questions are exercises in multiplication with assumptions you state out loud. Suppose you assume 10 million households, 20% of which buy a product twice a year: 10 million × 0.2 × 2 = 4 million purchases a year. These numbers are illustrations, not data; in an interview you would explain why each assumption is reasonable. What is usually being assessed is the structure and the arithmetic, and a clear structure with round numbers beats precise numbers you cannot justify. ## Percentages and margins Margins and shares are percentages of a base. Revenue of 250 and costs of 180 leave 70, and 70 out of 250 is 28%. To get there quickly, notice that 70 ÷ 250 is the same as 28 ÷ 100, since both numbers were multiplied by 0.4. Scaling a fraction to a denominator of 100 is often the fastest route to a percentage. Keep the benchmarks 1/8 = 12.5%, 1/6 ≈ 16.7%, 1/7 ≈ 14.3% and 1/3 ≈ 33.3% ready, because they turn many divisions into recognitions. ## Growth compounds Growth of 8% a year for three years is not 24%. It is 1.08 × 1.08 × 1.08, which is about 1.26, so roughly 26%. For small rates over a few periods, adding the rates gives a rough underestimate, and the gap widens as rates and periods grow. When precision matters, multiply step by step: 1.08 × 1.08 = 1.1664, and 1.1664 × 1.08 is about 1.26. ## The rule of 72, and its limits To estimate how long something takes to double at a constant growth rate, divide 72 by the percentage rate. At 6% a year, 72 ÷ 6 = 12 years; the exact figure is about 11.9. At 9%, the rule gives 8 years; the exact figure is about 8.04. It is an approximation derived from logarithms, and it becomes less accurate at high rates: at 24%, the rule says 3 years, while the exact doubling time is about 3.2. When you use it, say that it is an approximation. ## Break-even and per-unit thinking Break-even questions divide a fixed amount by a margin per unit. With fixed costs of 120,000 and a contribution of 8 per unit, break-even is 120,000 ÷ 8 = 15,000 units. Simplify before dividing: halve both numbers until the division is easy, 60,000 ÷ 4, then 30,000 ÷ 2. The same move, dividing both numbers by a common factor, speeds up most interview divisions. ## Talk through the calculation In a case interview, silence while you compute is harder to follow than a narrated calculation. Say what you are doing: I will round 31 to 30, so about 200, slightly less. Narrating lets the interviewer follow your reasoning and correct a misunderstanding early, and it gives you a structure to hold instead of a string of digits. Keep it short; you are showing the route, not reciting every addition. ## Write down what you are allowed to write If paper is permitted, write intermediate results down. Holding a number in your head that could be on the page is a working-memory cost with no benefit, and everything said above about pressure applies here in full. Where paper is not allowed, as in some online arithmetic tests, fall back on the methods with the fewest intermediate results. ## Build the fundamentals, then speed The preparation that pays off is unglamorous: fast, reliable single-digit facts, two-digit multiplication by splitting, division by simplifying, and common fractions as percentages. Practise in short sessions spread over the weeks before an interview rather than cramming the night before; Cepeda and colleagues found spaced practice was retained better than massed practice across many studies. Do some of it against a clock, since the interview will have one, even if nobody mentions it. ## Check before you commit Before announcing a result, run two checks: does the magnitude match your estimate, and do the units make sense? If you find an error, say so and correct it. Catching your own slip calmly is part of what careful work looks like, and it is far better than defending a wrong number. ## Where Numvolt fits Numvolt covers the fundamentals layer: addition, subtraction, multiplication and division. Zen mode, with no timer and the method behind every correction, is the place to make methods reliable, and timed sprints of 15, 30 or 60 seconds recreate a clock. It does not simulate a case interview, market sizing or finance questions; those need practice with realistic cases, ideally out loud with another person. ## Questions ### What math is in a consulting case interview? Mostly arithmetic with large numbers: multiplication, division, percentages and growth, usually combined with estimates you state and justify. The exact format depends on the firm, so check what it publishes about its own process. ### How can I calculate faster under interview pressure? Use methods that hold fewer numbers at once, such as separating the zeros and rounding before refining, and practise some sessions against a clock. Beilock and Carr linked pressure to a loss of the working memory available for the task. ### Is the rule of 72 exact? No. It approximates doubling time at a constant rate: at 6% it gives 12 years against an exact figure of about 11.9, and it drifts further at high rates. ### Can I use paper during an interview calculation? It depends on the format. When paper is allowed, write intermediate results down rather than holding them in your head. ## Further reading - Beilock & Carr (2005), When high-powered people fail: working memory and “choking under pressure” in math, Psychological Science — https://doi.org/10.1111/j.0956-7976.2005.00789.x - Ashcraft & Kirk (2001), The relationships among working memory, math anxiety, and performance, Journal of Experimental Psychology: General — https://doi.org/10.1037/0096-3445.130.2.224 - Cepeda et al. (2006), Distributed practice in verbal recall tasks: a review and quantitative synthesis, Psychological Bulletin — https://doi.org/10.1037/0033-2909.132.3.354