What this question is really asking
How many coffee shops are there in New York? How many umbrellas are sold in London each year? How many tennis balls fit in a Boeing 747?
These questions feel unfair the first time you hear one. You have no data, no internet, and about five minutes.
Here is the thing that makes them manageable: the interviewer does not know the answer either. And they do not care what number you land on.
What they are watching is whether you can break a huge unfamiliar problem into pieces that make sense, state your assumptions clearly enough to be challenged, do the arithmetic without losing your place, and notice when your answer is obviously wrong.
The worked example throughout this guide is coffee shops in New York City.
How to structure your answer
First, clarify the scope
Thirty seconds here prevents you answering a different question from the one asked.
Which geography exactly? New York City, or the metro area including New Jersey? Manhattan alone, or all five boroughs?
What counts? Independent cafés only, or chains too? Does a coffee counter inside a bookshop count?
Are we counting outlets, or estimating market size in money? These are completely different calculations.
Ask, then say the scope back so you both agree.
"So we are counting standalone coffee shops, including chains, across the five boroughs. Is that right?"
Choose your approach and name it
There are two ways to attack any sizing question.
Top-down
Start with a big number, usually a population, and narrow down.
Population of New York → how many drink coffee → how many buy it out rather than making it at home → how many cups per day → how many shops that supports.
Good when you know the population and can reason about behaviour.
Bottom-up
Start with one small unit and multiply up.
One coffee shop serves roughly X cups a day → each shop needs Y customers to survive → total demand divided by that gives the number of shops.
Good when you know something concrete about a single unit.
Most real answers use a bit of both. Say which one you are using and why.
Build the structure before touching any numbers
This is the single most important habit.
Say what you are going to multiply before you multiply it.
"I am going to start with the population, work out how many buy coffee out on a typical weekday, get to total daily cups, then divide by how many cups a single shop can sell in a day."
Two reasons this matters.
First, the structure is what is being marked. It is your thinking made visible.
Second, if one assumption turns out to be wrong, you can fix that number without the whole answer collapsing. If you have been calculating without a stated path, a wrong assumption means starting over.
Do the maths — the worked example
Now the numbers. Round hard, and say every step out loud.
Population. New York City has roughly 8 million residents. Round to 8 million.
Add commuters. Perhaps another 1.5 million people come in to work on a weekday. Call the daytime population 9.5 million. Round to 10 million for easier maths, and say that you are doing so.
Who buys coffee out? Not everyone. Children do not. Plenty of people make it at home. Say roughly half of the adult daytime population buys at least one coffee out on a weekday.
That is about 5 million cups. Some people buy two, so call it 6 million cups a day.
How many cups can one shop sell? A busy shop might sell 500 to 800 cups a day. A quieter one far fewer. Use 500 as an average.
Divide. 6 million ÷ 500 = 12,000 coffee shops.
Sanity-check it
This is the step almost nobody does, and it is one of the easiest ways to stand out.
Compare your answer to something you know.
12,000 shops across a population of 8 million residents is one shop per roughly 670 people. Does that feel plausible?
Think about a city block in Manhattan. It probably has one or two places selling coffee. Across the whole city including quieter residential areas, one per 670 people feels roughly right — maybe slightly high.
If your answer had come out at one shop per 50 people, you would say so and revisit the assumption causing it — probably cups per shop being far too low.
Catching your own error scores better than getting it right first time. It shows judgment rather than luck.
Give a final number with a range
Do not trail off. Commit.
"So I would say roughly 10,000 to 15,000, with a midpoint around 12,000. The number is most sensitive to how many cups a single shop sells per day — if that is 300 rather than 500, you get closer to 20,000."
Naming which assumption drives the answer is a strong finish. It shows you understand where the uncertainty actually lives.
Useful anchors to memorise
You do not need statistics. But a few round numbers make everything faster:
- United States: about 340 million people
- United Kingdom: about 68 million
- Germany: about 84 million
- New York City: about 8 million
- London: about 9 million
- Average household size in developed countries: about 2.5 people
- Life expectancy: about 80 years, useful for age-group questions
For age distributions, a rough rule: assume people are spread evenly across 80 years. So roughly a quarter are under 20, and roughly half are between 20 and 60. It is not precise, but it is defensible and fast.
Common question types
How many X are sold per year? Usually: population → how many own one → how often they replace it.
How many X are there in a city? Usually: demand → capacity per unit → divide.
What is the market size for X in money? Usually: units → price per unit → multiply.
How many X fit inside Y? Volume of Y ÷ volume of X, with a packing adjustment because objects do not stack perfectly.
Saying it out loud
"Let me first check the scope — five boroughs, all shops including chains, and we are counting outlets rather than revenue.
I will work from the daytime population down to daily cups, then divide by how many cups one shop can sell.
New York has about 8 million residents plus maybe 1.5 million commuters, so call it 10 million people in the city on a weekday.
If half of them buy a coffee out, and some buy two, that is roughly 6 million cups a day.
If an average shop sells 500 cups a day, that suggests about 12,000 coffee shops.
Sanity checking that, it is one shop per 670 residents, which feels about right for a dense city.
So I would say 10,000 to 15,000, and the answer is most sensitive to cups per shop."
What the interviewer is really checking
Structure over precision.
There is no correct answer. There are defensible structures and indefensible ones.
Can you do mental arithmetic out loud?
Under mild pressure, while explaining yourself. This is a learnable skill and it is being tested directly.
Do you state your assumptions?
Unstated assumptions cannot be challenged, and challenging one is exactly what the interviewer wants to do. It is how they test whether you can adjust.
Do you sanity-check?
Arriving at an implausible number and presenting it without comment is the clearest sign that you are processing rather than thinking.
Are you comfortable being roughly wrong?
The exercise rewards people who commit to an estimate and defend the reasoning. Hedging endlessly does not work here.
Frequently asked
Should I use top-down or bottom-up?
Whichever has the more reliable starting point. Top-down works when you know the population; bottom-up works when you know something about a single unit. Saying which you have chosen and why is itself part of the answer.
What if I do not know the population of the country or city?
Estimate it and say you are estimating. Learn a handful of anchors — US around 340 million, UK around 68 million, New York City around 8 million — and derive the rest. Nobody is testing recall of statistics.
How precise does the answer need to be?
Within an order of magnitude is fine. Interviewers are assessing structure and reasoning. An answer that is twice the true figure with clean logic beats a closer answer arrived at by guessing.
Can I ask the interviewer for data?
You can ask whether any figures are available, and sometimes they will give you one. Usually they will not, and the expectation is that you estimate and state the assumption. Do not stall waiting for numbers that are not coming.
What if my final number is obviously wrong?
Say so, identify which assumption is driving it, and adjust. Catching your own error is a strong signal — much better than presenting an implausible figure without comment, and often better than getting it right first time.
Common mistakes
Calculating before structuring.
Numbers without a plan is how candidates get lost halfway through.
Not stating assumptions.
If the interviewer cannot see where a number came from, they cannot give you credit for it.
Using awkward numbers.
8.34 million instead of 8 million makes every subsequent step harder for no benefit. Round and say you are rounding.
Going silent while calculating.
The reasoning is the answer. Silence gives the interviewer nothing to assess.
Never sanity-checking.
Presenting a figure implying one coffee shop per fifty people, without comment, is the worst outcome in this question type.
Chasing false precision.
Decimal places on an estimate built from guesses shows you have misunderstood the exercise.
Losing track of units.
Cups per day versus cups per year is by far the most common arithmetic error here. Say the unit every time.
Not committing to a number.
Structure with no conclusion is unfinished. Give an answer and a range.
Ready to practice?
Answer this question out loud with an AI interviewer.