Market Sizing in Case Interviews: A Step-by-Step Approach With a Worked Example

Market sizing questions ("How many coffee shops are there in this city?") sound like trivia, but they test structure and judgement: whether you can break a big, vague number into pieces you can reason about, make sensible assumptions out loud, and check that the answer makes sense. The exact figure is beside the point, since the interviewer doesn't know it either.

From the interviewer's chair, a sizing question is a five-minute window into how you think. I am listening for the order in which you do things: clarify, structure, estimate and then check, rather than reaching for a number in the first ten seconds and building a structure around it afterwards.

The steps

• Clarify what you're sizing: units or revenue, which geography, and what time period.
• Pause and lay out your approach before you use any numbers. Have the full equation in mind and say it out loud, for example: the number of coffee shops = (people who buy coffee out × cups each per day) ÷ (cups one shop sells per day).
• Split the problem into segments that add up to the whole without overlapping.
• State an assumption for each segment, rounded, with a reason.
• Calculate with round numbers, checking each step.
• Sanity-check the result against something you can reason about from a different starting point.
• Summarise the answer, and say what would change it.

The clarifying question takes fifteen seconds and prevents the most common disaster, which is sizing the wrong thing: cups when the interviewer wanted revenue, a country when they meant a city. Stating assumptions out loud matters because the interviewer scores the reasoning, and reasoning they cannot hear scores nothing. Say the number, say why, and say which direction you might be wrong in: "I'll take 40% of commuters as daily buyers, since many people make coffee at home or in the office, and if anything that may be high." A sentence like that invites a conversation rather than a correction.

Top-down or bottom-up

Top-down starts from a large known number, such as population, households or companies, and narrows it with percentages: who is relevant, who buys, and how often. Bottom-up starts from a single unit, such as one café, one store or one customer, and scales up: how much one unit produces, times how many units exist. Use one to build the estimate and the other as a cross-check. If the two land within a factor of two of each other, you have a defensible answer.

Choose the method that starts from the number you are most confident about. Consumer-spend questions tend to suit top-down, because population is a figure you know. Capacity and supply questions, such as how many petrol stations a country needs, tend to suit bottom-up, because you can picture the unit. Whichever you pick, name the choice and the reason before you begin, since that choice is itself a judgement the interviewer is scoring.

Worked example: annual takeaway coffee sales in a city of 2 million

Start by stating the full equation before you use any number: annual sales = working-age commuters × share who buy a takeaway coffee × cups each per working day × working days per year × average price. Then fill in each piece in turn.

• Start with working-age commuters. Assume half the population: 1,000,000 people.
• Assume 40% buy a takeaway coffee on a working day: 400,000 buyers.
• Assume 1.2 cups each per day: 480,000 cups per working day.
• Assume 250 working days a year: 480,000 × 250 = 120 million cups.
• Assume an average price of $3.50: 120 million × $3.50 = $420 million a year.

Now cross-check bottom-up. A busy café might sell around $600,000 of takeaway coffee a year. $420 million ÷ $600,000 = 700 cafés, which is about one café per 2,900 residents. That's plausible for a city, so the estimate is in a sensible range. Finish by saying what you left out, such as weekends, tourists and office or vending machine coffee, and which way each would move the number.

Second worked example: cups sold by coffee shops in a city of 2 million, bottom-up

Now take the same city from the other direction, starting with a single shop and scaling up to find how many cups the city's coffee shops sell in a year. The equation, stated before any number: annual cups = number of coffee shops × cups one shop sells per day × days open per year.

• Number of shops. Picture a neighbourhood of 10,000 residents: a high street with three or four cafés is a familiar scene, so take 4 per 10,000 people. The city has 2,000,000 ÷ 10,000 = 200 such neighbourhoods, so 200 × 4 = 800 coffee shops.
• Cups per shop per day. Split the day into peak and off-peak. In the two-hour morning rush a counter serves about one cup a minute: 120 cups. For the remaining eight hours assume 15 cups an hour: 8 × 15 = 120 cups. Total: 240 cups a day.
• Days open. Allowing for holidays and quiet Sundays, assume 300 days a year. 240 × 300 = 72,000 cups per shop per year.
• Scale up. 800 shops × 72,000 cups = 57.6 million cups a year. Say about 60 million.

Cross-check it two ways. Per head, 57.6 million cups ÷ 2 million residents is about 29 cups per resident per year, or one a fortnight, believable for a commuter city. Against the top-down example above, which gave 120 million takeaway cups, the bottom-up figure is about half, inside the factor-of-two range, and the gap has a reason you can name: the top-down number counted takeaway coffee from any outlet, including kiosks, bakeries and petrol stations, while this one counts dedicated coffee shops only. Said out loud, that turns a discrepancy into evidence that you understand what each number measures. Had the two come out ten times apart, you would go back and find the assumption that broke.

Sanity-checking against a figure you know

The sanity check is the step candidates skip when the clock is running, and it is the one that separates a number from an answer. Convert your result into a unit you have intuition about and ask whether it feels right: per person per year, per household per week, per shop per day, or share of a country's GDP. In the first example, $420 million across 2 million residents is $210 a head a year, or about $4 a week, plausible for a city where many people commute. Keep a short list of reference figures in your head: the population of your country and its three largest cities, a household size of two to three people, 250 working days and 52 weeks in a year, 8,760 hours in a year, and a life expectancy near 80. Most sizing questions can be anchored to one of those.

When the interviewer challenges an assumption

Expect at least one push. A typical line is: "I'd have put that nearer 20% than 40%." The challenge is seldom about the number; it tests whether you tracked your own chain of logic and can take new information without losing your footing. Hold your ground in one sentence if you have a reason, then offer to run the alternative. Ask whether the interviewer has a figure in mind, and use it if they do, because a number from the interviewer beats a guess from you. Then redo the arithmetic out loud: 20% instead of 40% halves the buyers to 200,000, so 240,000 cups a working day, 60 million cups a year and about $210 million. Two responses fail: folding without a word, which tells me the original assumption had no reasoning behind it, and defending a guess as if it were data.

What a strong answer sounds like

Spoken, the first example runs like this, in under three minutes: "Before I touch numbers, my equation is annual sales equals commuters, times the share who buy, times cups a day, times working days, times price. Of 2 million residents I'll take half as working-age commuters, so 1 million. I'd assume 40% buy a takeaway coffee on a working day, since many people make it at home or at the office, which gives 400,000 buyers. At 1.2 cups each that is 480,000 cups a day, and over 250 working days, 120 million cups. At $3.50 a cup, that comes to about $420 million a year. Before I rely on that, I'd cross-check it from the café side."

Where candidates go wrong

• Diving into numbers before laying out the structure. Fix: write the equation down and read it aloud before the first assumption.
• Segments that overlap or leave gaps. Fix: ask whether a given customer could sit in two segments, or in none.
• Assumptions with no reasoning ("let's say 30%"). Give a reason, even a rough one, and say which direction you are more likely to be wrong in.
• False precision. Say about $420 million, not $419,872,000. Round at each step and give two significant figures at the end.
• Skipping the sanity check, or the "so what": what the number means for the client's decision. Fix: convert the result to a per-head figure and say in one sentence what it implies.
• Losing the thread when a figure changes. Fix: keep a running column of intermediate results on your page, with a label beside each one, so you can recompute in seconds.

How interviewers score this

Sizing touches all five scored skills, which is why it appears so often. Structuring is the equation and the segmentation, stated before any numbers. Quantitative analysis is the arithmetic, done with round figures, and the speed at which you recover when a slip appears. Business judgement shows in the assumptions: 40% of commuters buying coffee is a judgement, and the reason you give for it is what gets scored. Synthesis and recommendation is the last minute, where the number becomes a statement about the client's decision: a $420 million market is large enough to enter, or too small to justify a second roastery. Communication is whether I could follow you without looking at your paper. A candidate 30% off with clear reasoning outscores one who hits the right figure by a route I could not follow.

How to practise

Sizing is a skill you build by doing many small ones out loud: five minutes, a timer, a different market each day. Then practise it inside full cases, where sizing is often the first step toward a decision. CaseClimb includes sizing cases in its case library, so you can practise this within a realistic interview.

Common questions

How long should a market-sizing answer take? Three to five minutes from clarifying question to summary for a standalone sizing question. Inside a longer case, where sizing is one step towards a decision, aim for two to three minutes. Past five minutes the interviewer tends to move you on.

Does the interviewer know the real answer? Sometimes, and it makes no difference to your score. The interviewer is marking structure, reasoning and the sanity check, so an estimate that is 40% off with sound logic scores well. A figure from memory with no working behind it scores low even when right.

Is it acceptable to change an assumption halfway through? Yes, if you say so and say why. Noticing that your per-shop figure implies 2,000 cups a day, which no counter could serve, and correcting it, is the sanity check working as intended. Changing figures without comment, or steering them towards an answer you had in mind, is what costs marks.

Can I practise market sizing on my own? Yes. Pick an object in the room, set a five-minute timer, and size its market out loud, then write down the answer and the assumption you were least sure of. Ten minutes of timed mental maths a day keeps the arithmetic from slowing you down, and a structured prep plan shows where sizing drills fit alongside full cases.