How to do mental maths and market sizing under pressure
In a case interview you do arithmetic aloud, with a person watching, and you often have to estimate a number nobody gave you. Accuracy matters, but so do method, pace and a sensible answer. This guide covers the habits that make arithmetic faster and safer, then a simple approach to market sizing with a worked example using invented figures. Interviewers differ on whether a calculator or paper is allowed, so ask at the start.
How do I make the arithmetic itself faster?
Break numbers into easy blocks and work with round figures.
- Percentages. 10% is a decimal shift, 5% is half of that, 1% is a hundredth. So 17.5% of 240 is 24 plus 12 plus 6, which is 42. Check: 10% is 24, 5% is 12, and 2.5% is 6.
- Fractions that recur. 12.5% is one eighth, 16.7% one sixth, 25% one quarter, 33.3% one third, 37.5% three eighths.
- Percentage change uses the starting figure as base. A fall from 80 to 68 is 12 / 80 = 15%. Recovering needs 12 / 68, about 17.6%, because the base is smaller.
- Percentage points versus percent. A margin moving from 12% to 15% is 3 points, and a 25% relative increase. Say which you mean.
- The rule of 72. Years to double is roughly 72 ÷ the growth rate in percent. At 6% it is about 12 years. It is an approximation that works best for modest rates.
- Large numbers. Convert to a unit and keep the unit with you: 1.2m times 2.5 is 3.0m. Write units next to every figure, and keep millions, billions and thousands apart.
- Multiplying. Split one number: 36 times 25 is 36 times 100 ÷ 4, which is 900.
How should I do a calculation aloud?
Say the method, then the figures, then the answer with its unit, then a quick check on size. For example: "Revenue is 120 customers times 5 visits times £8. 120 times 5 is 600, times 8 is £4,800, so around 5 thousand. That seems plausible for a small shop." If you slip, say so and correct from the point of the error. Interviewers generally value a calm correction more than a silent wrong answer.
When can I round, and how much?
Round when the inputs are themselves uncertain. If one input is "about 2.4 people per household", rounding to 2.5 or 2 loses less than the uncertainty already there. Round in a direction that will partly cancel: one up, one down. Keep more precision where small differences decide the answer, such as a break-even or a choice between two close options. State your rounding aloud so the interviewer can follow it.
How do I approach a market-sizing question?
Choose a method before you calculate, and say why.
- Top-down starts from a total you can anchor (population, households, an industry total) and narrows it by shares. It is quick and suits a slice of a known total.
- Bottom-up builds from units: how many customers, how much each buys, at what price. It suits a new product or business market and shows which assumptions matter.
Then follow four steps: clarify what is being measured (units or value, which year, which geography), segment the population into groups that do not overlap and differ in how much they buy, state each assumption aloud with round numbers, and finish by saying which segment carries most of the answer. Where you can, size the market a second way and compare.
What does a worked example look like?
Question: roughly how much do residents of Aldenmoor, a city of 1.2m people, spend each year on servicing electric cargo bikes? All figures are invented assumptions.
Top-down, computed before writing:
- Households: 1.2m / 2.4 people per household = 500,000.
- Households owning a cargo bike: 4% of 500,000 = 20,000 bikes.
- Services per bike per year: 2, giving 40,000 services.
- Price per service: £60, giving 40,000 x 60 = £2.4m a year.
Cross-check, bottom-up from the supply side: if there are about 25 repair shops in the city that each do around 1,600 cargo-bike services a year, that is 40,000 services, or again £2.4m. The two routes agree, but both rest on guesses, so I would say the answer is "a little over £2 million, in a range of perhaps 1.5 to 3 million", and name which assumption I trust least, here the 4% ownership figure.
What to say aloud: "I'll size this top-down by households and cross-check from supply. There are about 500 thousand households. I assume 4% own a cargo bike, so 20 thousand bikes, serviced twice a year at £60, which gives around £2.4 million. As a check, 25 shops at 1,600 services each gives the same volume. The assumption I'm least sure of is ownership, so if I could ask for one piece of data it would be that."
What if I realise an assumption is wrong?
Say so, correct it, and update the answer. "I assumed 2.4 people per household; if the city skews towards students it may be nearer 3, which lowers households by a fifth and the answer with it." A candidate who recovers openly usually does better than one who defends a bad number. If the interviewer gives you an anchor, use theirs.
How do I sanity-check?
Compare to a known anchor: per-person spend, share of a bigger market, or a plausible number of businesses. Check units and zeros. Ask whether the answer is the right order of magnitude. A bike-servicing market larger than the city's whole retail sector would fail the check at once.
What mistakes should I avoid?
- Starting to calculate before choosing and stating a method.
- Segments that overlap or miss people.
- Silent arithmetic.
- False precision, such as £2,413,788 from guessed inputs.
- Mixing units or periods (monthly and yearly).
- Forgetting to say what you would do with the number.
What to do this week
- Do ten percentage-of-a-number sums aloud with no calculator, in under ten seconds each.
- Memorise the recurring fractions and the rule of 72.
- Size two everyday markets, once top-down and once bottom-up, and compare.
- Write down five anchor numbers you would be happy to quote and check them from reliable public sources.
- Record yourself on one estimate and listen for silence and unstated assumptions.
About this guide: it is original to Charta Learning and has not yet been reviewed by an experienced practitioner. Names in examples are invented. All figures are invented assumptions for practice, not data, and interviewers differ on whether paper or a calculator is allowed.