In this article
Before you start
Why interviewers ask this
A sizing question shows, in five minutes, whether you can turn a vague business question into a structure and defend each number in it. Nobody in the room knows the exact answer, so the only thing to grade is your reasoning.
Boston Consulting Group (BCG) says of its case interviews that "what matters most is how you approach the problem and the quality of your reasoning", and lists structuring an approach, asking good questions and doing quick calculations among the things candidates do (BCG careers, read October 2026). Bain describes case interviews as roles where you "work through a problem" (Bain careers, read October 2026). A weak answer recites a memorised number; a strong one is a small model the interviewer can change.
Market sizing from first principles
- Market sizing
- Market sizing is an estimation method that produces the number of units, customers or revenue in a defined market by breaking it into smaller quantities that can be reasoned about, multiplying them, and checking the result against a second route.
- Driver tree
- A driver tree is a diagram that splits one quantity into the smaller quantities that multiply or add up to it, so each branch can be estimated and challenged on its own.
- Triangulation
- Triangulation is the practice of estimating the same quantity by two independent routes and comparing them, so that agreement raises confidence and disagreement shows which assumption to revisit.
In plain words
Market sizing works like guessing how many people a stadium holds. You do not guess the total. You count the sections, guess the rows in one section, guess the seats in one row, and multiply. Each guess is small, so it is reasonable, and the errors partly cancel. The analogy breaks in one place: a stadium has a true answer you can look up, while a market often has only definitions, so you must say exactly what you are counting. Stated precisely, a market sizing is a driver tree. It has rounded numbers on its leaves, a stated range and a second route that checks it.
Step 1: Clarify what you are sizing
Clarifying fixes the unit, the place, the time period and the meaning of "market" before any number appears. The input is the interviewer's one-line question; the output is a sentence such as "yearly units sold in the United States, not revenue and not worldwide". This is the common error: candidates start multiplying without agreeing whether "market" means units, customers or revenue, and a correct calculation of the wrong quantity answers a question nobody asked. Ask two or three short questions, state your definition aloud, and wait for a nod. A definition also tells you which driver tree to build. What you now know: the quantity, its unit and its boundary.
Step 2: Choose an approach, top-down or bottom-up
A top-down estimate is a figure that starts from a large known total and keeps only the share that applies; a bottom-up estimate is a figure that starts from one buyer or one outlet and scales up. The input is your clarified definition; the output is one route, plus a second you will use to check. Top-down is quick when a reliable total exists, such as a count of households. Bottom-up is easier to defend because every factor is a physical thing you can picture. A strong answer picks one as the main route and names the other as the cross-check. What you now know: which route carries the answer and which one verifies it.
Step 3: Build the driver tree and state assumptions
The driver tree writes the quantity as a product of factors, and each leaf gets a rounded number with a reason. The input is the chosen route; the output is a tree of three to five factors you can say in one breath, for example households, share that buy, units per buyer per year. Say each assumption out loud as you write it, because a number stated as an assumption can be changed by the interviewer, while a number presented as fact can only be challenged. Prefer numbers you can anchor to something known, such as a count of households or a share of the population. What you now know: every number has an owner and a reason.
Step 4: Calculate with round numbers
Calculating means multiplying the leaves with figures rounded so that the mental arithmetic stays easy, and writing the units next to each number. The input is the tree; the output is a base figure and, if time allows, a low and a high case. Round the 134.79 million Census households of the worked example (Census household series, updated December 2025) to about 135 million, multiply in millions, and keep units attached so that a slip shows up as a wrong unit. Say the intermediate result at each multiplication, so the interviewer can follow every line. If a number surprises you, pause and look at it before continuing. What you now know: a single base figure with its units.
Step 5: Cross-check with a second route
A cross-check estimates the same quantity by an independent route and compares the two figures. The input is your base figure; the output is a stated range and a sentence about how well the two routes agree. If they differ by less than a factor of two, report the overlap; if they differ by more, find which assumption is the least certain and say so. This is the step that turns a calculation into an answer an interviewer can trust, and it is the one most often skipped when time runs short, so plan for it from the start. What you now know: how far to trust the number.
Step-by-step description
- Fix the unit, place and period.
- Top-down or bottom-up, plus a check.
- Split into factors with stated assumptions.
- Multiply rounded numbers with units.
- Compare with a second route, state a range.
- Clarifying fixes the unit, the place, the time period and the meaning of market before any number appears.
- A cross-check estimates the same quantity by an independent route and compares the two figures.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →Worked example
The question is "How many replacement cartridges for countertop water-filter pitchers are sold in the United States each year?" Step 1 fixes the unit as cartridges per year in the United States. Only one input is sourced: the Census Bureau's Current Population Survey (CPS) counts 134.79 million households in 2025 (U.S. Census Bureau via Federal Reserve Economic Data, updated December 2025). Every other number below is illustrative, chosen to show the method and not measured.1
Route A is top-down (demand side, from the household total): households, times the share that own a pitcher filter, times the cartridges each owner replaces per year. With illustrative values of 20%, 25% and 30% ownership and 4, 6 and 8 cartridges a year, the estimate is 107.8, 202.2 and 323.5 million cartridges a year (low, base, high).1
Route B is bottom-up (supply side, from one store scaled up): stores that stock cartridges, times the cartridges each sells per week, times 52 weeks. With an illustrative 50,000 stores selling 60, 100 and 140 a week, the estimate is 156.0, 260.0 and 364.0 million.1
The two bases differ by 28.6%, and the ranges overlap from 156 to 323.5 million, so the stated answer is "roughly 150 to 300 million cartridges a year, most likely around 230 million", where 229.3 million is the geometric mean of the two bases (the square root of the two multiplied together, a midpoint suited to numbers that differ by a ratio).1 The sourced input can be checked too. The Census Bureau reports 39.7 million one-person households, 29% of all households (U.S. Census Bureau, December 2025); 39.7 divided by 134.79 is 29.5%, which agrees with the rounded share.1
| Fact | Value |
|---|---|
| US households, 2025 (sourced) | 134.79 million |
| Route A, demand side, low / base / high (illustrative) | 107.8 / 202.2 / 323.5 million a year |
| Route B, supply side, low / base / high (illustrative) | 156.0 / 260.0 / 364.0 million a year |
| Gap between the two bases | 28.6% |
| Overlap of the two ranges | 156.0 to 323.5 million a year |
| Route A base if ownership were half as high | 101.1 million a year |
Text description of the diagram
- The ranges overlap from 156 to 323.5 million cartridges a year.
- The two bases differ by 28.6%.
Trade-offs and edge cases
The textbook method works when a market has a clear buyer and a countable unit, and it bends in three places. First, a market that does not exist yet has no ownership rate to look up, so you size an analogous market and adjust it, saying which way and why.
Second, a business-to-business market has no population to start from; count the firms in the target segment, then the buyers inside each firm, then the budget or the units each buyer owns. Third, revenue needs one more factor than units, the price, and price is usually the weakest number. How much one assumption moves the answer is worth stating: halving the ownership share halves route A, from 202.2 to 101.1 million.1
Text description of the diagram
- A top-down estimate starts from a large known total and keeps only the share that applies.
- A bottom-up estimate starts from one buyer or one outlet and scales up.
Follow the interviewer on precision: BCG notes that there is not always a single right answer, which is why the stated range and the reasoning carry the score, not the last digit (BCG careers, read October 2026). A year-dated population figure also ages: the household count above is the 2025 value, and a later release will replace it.
How to answer it in an interview
The spoken answer follows the five steps in about a minute and a half: define, choose, name the drivers, give the number, check it. Say the structure before the numbers, so the interviewer knows where you are going.
How many replacement cartridges for water-filter pitchers are sold in the United States each year?
Let me confirm: units per year, United States, replacement cartridges only. I will size it top-down from households and check it bottom-up from stores. Demand is households, times the share with a pitcher filter, times cartridges per household a year. About 135 million households, say a quarter own one, six cartridges each, gives roughly 200 million. For the check, stores: about 50,000 outlets selling around 100 a week is 260 million. The routes are within thirty percent, so I would say 150 to 300 million, and ownership is the assumption I would test first.
The spoken structure in one sentence: say the definition, say the route, give the number with its range, then name the weakest assumption.
Common mistakes
- Starting with arithmetic. (wrong) "So, 300 million people, divided by four..." It skips the definition. (right) "Let me confirm we mean units a year in the United States, not revenue."
- Presenting guesses as facts. (wrong) "A quarter of households own one." (right) "I will assume a quarter own one, based on how common these are; tell me if you want a different figure."
- Using one route only. (wrong) A single product of factors with no check. (right) "As a cross-check, I will go from stores to units."
- Over-precise answers. (wrong) "The market is 202,185,000 cartridges." (right) "Around 200 million, within a range of 150 to 300 million."
- Silent arithmetic. (wrong) Thirty seconds of quiet, then a number. (right) Say each intermediate result, so a slip is caught early. The same habit helps in behavioral interviews, where structure is also graded.
Follow-up questions
Is top-down or bottom-up better for market sizing?
Neither wins everywhere: use the route whose inputs you can defend, and the other as the check. Top-down suits a market with a reliable total, bottom-up suits a new product or a small market. The probability and statistics hub covers the estimation habits behind both.
What if I do not know an input number?
Say that you do not know it, give a round figure with a reason, and offer a range. A stated assumption the interviewer can adjust is better than a hidden one, and interviewers often supply the figure on request.
How accurate does a market-sizing answer need to be?
As a working rule on this page, a cross-check within a factor of two is close enough to report as one range; it is not a stated interviewer tolerance, because the structure and the check are what carry the score. Report a range, and name which assumption would move it most; the worked example shows that halving one factor halves the answer.1
How do you size a market that does not exist yet?
Size the analogous market, then adjust for how the new product differs: its price, its audience and how quickly people would adopt it. Name the analogy so the interviewer can challenge it. Finance questions use the same logic when they value a business from comparables, as in the investment banking interview guide, and the interview loop overview shows where case rounds sit. Product candidates meet the same estimates in the product management interview hub, and tell me about yourself is the opener that usually comes before the case.
- Define the unit, place and period before any number.
- Pick one route as the answer and one as the check.
- State every assumption aloud, rounded, with a reason.
- Give a range, then name the assumption you would test first.
- Two independent routes within a factor of two earn trust.
Sources
- Total Households (TTLHH), U.S. Census Bureau series on FRED — Federal Reserve Bank of St. Louis (U.S. Census Bureau data), 4 Dec 2025
- Census Bureau Releases New Estimates on America's Families and Living Arrangements (CB25-TPS.78) — U.S. Census Bureau, 2 Dec 2025
- Case Interview Preparation — Boston Consulting Group careers site (read 2026-10-24), 24 Oct 2026
- Our Hiring Process — Bain & Company careers site (read 2026-10-24), 24 Oct 2026
