Market sizing is the part of a case interview that looks easiest and quietly costs the most points. Nobody expects you to know how many gas stations there are in the United States. The interviewer wants to watch you build a number out of parts you can defend, out loud, in about three minutes, and then notice whether the number makes sense.
That is a learnable skill, and the way to learn it is to work examples end to end. Below are six, from a consumer product to a reverse breakeven, each with every assumption stated so you can disagree with it. They start from published numbers where they can, such as the Census Bureau’s population estimate, and one is checked against a count published by the convenience store industry, so you can see how close a careful two-minute estimate gets.
First, the method the examples follow.
Two ways into the same number
Every market sizing answer starts from one of two ends. Top-down starts from a population and narrows it: people, then the share who use the thing, then how often they buy it. Bottom-up starts from a unit you can picture and multiplies: one store, how many customers it serves an hour, how many hours it is open, how many stores.
Top-down works when demand is easy to reason about, as with toothbrushes or streaming subscriptions. Bottom-up works when supply is easier to picture than demand, as with a barbershop, a factory line or a fleet of trucks. The strongest answers use one to estimate and the other to check, and saying that out loud (“let me check this from the supply side”) is one of the clearest signals you can send.
The five steps interviewers listen for
The method matters more than the answer, so it is worth making it a habit. These five steps take about three minutes once they are automatic.
1. Clarify what you are sizing. Units or revenue? Which country, which year? Do replacements count, or only first purchases? Is it the installed base (how many exist) or the annual flow (how many are sold)? One clarifying question here often changes the answer by a third, as the toothbrush example below shows.
2. Lay out the structure before any numbers. Say the formula first: “people, times the share who use a manual brush, times how many they buy a year.” The interviewer can follow, and agree or redirect, before you have spent any time on arithmetic.
3. State each assumption with a round number and a reason. “About 342 million people, from the latest Census estimate.” “Say 70 percent use a manual brush, since electric brushes are common but not the norm.” Round numbers keep the math fast; the reason shows the number is not a guess.
4. Calculate out loud. Say each step, including the units. Interviewers score arithmetic, and a silent pause followed by a number gives them nothing to score.
5. Sanity-check, then say the so-what. Check the answer a second way, against a fact you know, or against capacity. Then say what it means for the question behind the question: “so the market is big enough that a 2 percent share would cover the plant.”
- 1ClarifyUnits or revenue, geography, year, installed base or annual sales.
- 2Structure firstSay the formula before any number.
- 3Assume out loudRound numbers, one reason each.
- 4Calculate out loudEvery step, with its units.
- 5Check, then so-whatA second method or a known fact, then what the number means.
1. How many toothbrushes do Americans buy a year?
Clarify first. Does an electric toothbrush count? What about replacement heads? Ask, because the answer moves a lot. Say the interviewer wants manual toothbrushes and electric handles, and is curious about heads.
Structure. Population, split by manual and electric users, times how often each group buys.
Assumptions. The Census Bureau puts the US population at about 342 million as of July 2025. Say 70 percent brush with a manual toothbrush and 30 percent with an electric one. The American Dental Association recommends replacing a brush every three to four months, which would be three or four a year, but plenty of people stretch it, so assume manual users actually buy 2.5 a year. Electric users replace the handle every three years or so and the head twice a year.
Math. Manual: 342 million times 0.7 is about 239 million people, times 2.5 is about 600 million toothbrushes a year. Electric handles: 342 million times 0.3 is about 103 million people, divided by 3 is about 34 million handles. Heads: 103 million times 2 is about 205 million.
Check and so-what. About 600 million manual brushes is just under two per person per year, which feels right for a habit most people follow loosely. The useful observation for the interviewer is the definition question: counting heads adds a third to the unit market, and a client selling electric brushes ships about six heads for every handle.
2. How many gas stations are there in the United States?
This one is worth doing from demand and from supply at once, because neither end is obvious.
Structure. Fill-ups needed per day, divided by fill-ups one station can handle per day.
Demand. About 342 million people. Say 0.85 vehicles per person, about 290 million vehicles. Each fills up about once a week, so 290 million times 52 is about 15 billion fill-ups a year, or about 41 million a day.
Supply. Picture a typical station: 8 pumps, open 18 hours. A fill-up takes about six minutes including pulling in and out, so a pump can handle 10 an hour at most. Stations are busy at rush hour and empty at 3pm, so assume 25 percent utilization across the day. That is 8 times 18 times 10 times 0.25, or 360 fill-ups a day.
Math. 41 million divided by 360 is about 115,000 stations.
Check. The convenience store industry association, NACS, counted 122,620 convenience stores selling fuel at the end of 2025, and says convenience stores sell about 80 percent of the fuel Americans buy. Fuel stations that are not convenience stores are on top of that, so the real total is somewhat above 122,620. An estimate of 115,000 is in the right range and a little low, probably because many stations have fewer than 8 pumps. That is exactly the kind of sentence to say out loud.
3. How many haircuts can one barbershop do in a year?
A capacity question, so bottom-up is the natural way in. It is also a common first step inside a case: Busch’s Barber Shop, from the Kellogg casebook, sizes a barbershop’s market before it gets to the location economics.
Structure. Chairs, times hours each chair is open, times cuts per hour, times how busy the chairs are.
Assumptions. Four chairs. Open six days a week for 10 hours, 50 weeks a year, which is 3,000 hours per chair. A cut takes about 30 minutes with cleanup, so two an hour. Chairs are full about 60 percent of the time, busier on evenings and Saturdays.
Math. 4 chairs times 3,000 hours is 12,000 chair-hours. Times 2 cuts is 24,000, times 0.6 is 14,400 haircuts a year. At a $30 average ticket, that is about $432,000 in revenue.
Check from demand. If a regular customer comes every five weeks or so, about 10 times a year, the shop needs about 1,440 regulars. In a town of 75,000 people with roughly 37,500 men, that is about 4 percent of them, which is plausible for one well-located shop and tells you where the risk is: the shop lives or dies on winning one in 25 local men.
4. What does a coffee kiosk in one airport terminal earn in a year?
A revenue question that mixes a population (passengers) with a capture rate. Booth’s Burger Palace case runs a version of this for a burger chain at Chicago’s Midway airport, built up from gates and load factors.
Structure. Passengers who walk past, times the share who buy, times the average ticket, plus airport staff.
Assumptions. The terminal handles 10 million departing passengers a year, and 40 percent of them pass this kiosk on the way to their gates, so 4 million. Five percent buy something: most have already eaten, are rushing, or buy from the bigger chain near security. That is 200,000 transactions. Add airport staff: 2,000 people work near the kiosk, on about 200 days each, and 30 percent of them buy on a given day, another 120,000. Average ticket: $7.
Math. 320,000 transactions times $7 is about $2.2 million a year.
Check against capacity. 320,000 transactions a year is about 880 a day. Over 18 open hours, that is about 49 an hour, or one every 75 seconds. That is busy but doable with two baristas, so the estimate does not need more staff than the kiosk could hold. If the math had said one sale every 15 seconds, the capture rate would be the assumption to revisit.
5. How many public EV chargers does a city of a million people need?
An infrastructure question where the answer depends on a choice you should make explicit. Wharton’s Electro Chargers case asks a related question: which city a battery-swap company should enter, and what the network would cost.
Structure. Electric cars, times the energy they need each day, times the share charged in public, divided by what one public charger delivers in a day.
Assumptions. A dense city has about 0.7 cars per person, so 700,000 cars. Say 5 percent are electric, 35,000 EVs. An average car drives 30 miles a day at about 0.3 kWh a mile, so 9 kWh a day. Most charging happens at home; say 30 percent happens on public chargers.
Math. 35,000 cars times 9 kWh is 315,000 kWh a day. Thirty percent of that is about 94,500 kWh a day on public chargers. Now the choice. A Level 2 charger delivers about 7 kW and is busy maybe 8 hours a day, so about 56 kWh a day: you would need about 1,700 of them. A 50 kW DC fast charger busy 4 hours a day delivers 200 kWh: you would need about 470.
So-what. Same demand, two answers that differ by a factor of almost four. The right answer depends on where people park: Level 2 at workplaces and apartments where cars sit for hours, fast chargers along roads where they do not. Saying that is worth more than either number.
6. How many members does a new gym need to break even?
A reverse sizing question: instead of estimating a market, you work out the share of it the client must win, then judge whether that is realistic. Burger Corp, from The Ultimate Case Interview Workbook, uses a reverse sizing step like this inside an acquisition case.
Structure. Fixed costs, divided by what each member contributes, then compared with the local market.
Assumptions. Rent, staff and equipment leases come to $60,000 a month. Membership is $50 a month, and each member costs about $5 a month in towels, cleaning and wear, so each contributes $45.
Math. $60,000 divided by $45 is about 1,333 members to break even.
Check, twice. Capacity first: if 30 percent of members come on a given day, that is 400 visits, about 27 an hour over 15 hours, which a mid-size gym floor can hold. Then the market: a town with 50,000 adults where about 20 percent hold a gym membership has 10,000 members across all gyms, so this gym needs about 13 percent of them. Whether that is realistic depends on how many competitors there are, which is the next question to ask the interviewer.
The mistakes that cost points
Most points in market sizing are lost to the same handful of habits, and none of them are about arithmetic ability.
Skipping the definition. Sizing “the toothbrush market” without asking whether heads count, or “the gym market” without asking whether boutique studios do. One question at the start saves a rework at the end.
Mixing stock and flow. The number of cars on the road is a stock; the number sold this year is a flow. Multiplying a stock by a purchase rate, or comparing a stock with a flow, is the most common structural error.
False precision. “There are 114,722 gas stations” tells the interviewer you have stopped thinking. Round numbers, and say the answer as a range or with “about”.
Silent math. A pause and then a number gives the interviewer nothing to score, and if the number is wrong they cannot see where. Say each step.
No sanity check. The answer that most impresses is the one where the candidate says “that looks high, let me check it from the supply side” before the interviewer does.
No so-what. The number is never the point. The point is what it means for the client: big enough to enter, too small to justify the plant, dependent on winning one in 25 local customers.
Practice it out loud, inside a full case
Standalone estimates are good warm-ups, but in an interview market sizing usually sits inside a case, and the number you produce feeds the next calculation. A shaky estimate in minute eight becomes a wrong breakeven in minute twenty. That is why it pays to practice sizing inside full cases with an interviewer who asks why after every assumption.
Several of the 38 interviewer-led cases on Tough Tongue AI have a market sizing step at their core: Burger Palace (customers from airport gates and load factors), Busch’s Barber Shop (a barbershop’s local market), Pink Cross (a segmented sizing estimate), Telcast (a capacity estimate for broadband base stations) and Burger Corp (a reverse sizing problem). The interviewer waits for your structure before giving data, questions the assumptions that matter, makes you find your own arithmetic errors, and grades the result against the case’s answer key.
The case interview prep hub lists every case by casebook and difficulty. If your interviews are at McKinsey, the McKinsey practice cases page maps each question type, sizing included, to the cases that drill it. And if you would rather start for free, the guide to practicing case interviews by yourself has a ChatGPT and Claude prompt you can point at any of the six questions above.
