Market Sizing: A Step-by-Step Method (with Examples)

Market sizing is the case interview skill of estimating the total size of a market when you have no exact data. You crack it in four steps: pick a top-down or bottom-up approach, break the population into segments with a tree, choose sane assumptions and state them, then multiply through and sanity-check the result. The number rarely matters as much as the structure and the judgment behind your assumptions. Here is the method, with two fully worked examples.

When I interviewed at McKinsey, I never expected a candidate to know the true number of gas stations in the United States. I expected a clean tree, defensible assumptions said out loud, and arithmetic that didn't lose a zero. Get those three right and you pass, even if your estimate is off by a factor of two. Get them wrong and a "correct" number won't save you.

Top-down vs bottom-up: pick your direction

Every market-sizing question is approached one of two ways. Choosing the right one, and saying why, is the first thing the interviewer grades.

Top-down starts wide. To size the market for running shoes, you begin with a country's population and whittle it down by who runs, how often they buy, and what they pay. Bottom-up starts narrow. To size a coffee shop's annual revenue, you estimate one store's daily sales and scale by the number of stores. Strong candidates often run one approach, then cross-check with the other. If both land in the same ballpark, your confidence (and the interviewer's) goes up.

The segmentation tree

The tree is the backbone of market sizing. You take a big quantity and split it into branches you can estimate, multiplying along each path. The discipline is to split into pieces that behave differently, not pieces that are just smaller.

A good tree for a consumer product looks like this:

• Total population
• times the share in the relevant age or demographic segment
• times the share who are actual buyers
• times purchase frequency per year
• times average price paid

Each branch is a number you can defend with one sentence. The goal is three or four clean multiplications, not fifteen tiny ones. Every extra branch is another place to lose a zero or compound an error, so keep the tree shallow and the assumptions honest.

Choosing sane assumptions

This is where most candidates either shine or sink. An assumption is sane when you can justify it in one sentence and it sits in a believable range. A few principles I drilled into candidates:

• Anchor to numbers you actually know. Memorize a handful of reference figures: a country's rough population, an average household size, a typical workweek. Build outward from those anchors instead of guessing into the void.
• Round to friendly numbers as you go. Use a population of 330 million, not 331,449,281. Clean inputs keep the arithmetic fast and the focus on logic.
• Split where behavior differs. Segment urban from rural, or daily from occasional users, only when those groups buy at genuinely different rates. Otherwise you've added complexity for nothing.
• Say every assumption out loud. The interviewer is grading your judgment, not your psychic powers. "I'll assume 60 percent of adults drink coffee, since it skews high in most developed markets" tells them how you think.

A useful set of anchors to walk in with:

One more thing about anchors: you don't need many of them. Five or six well-chosen reference figures let you build almost any consumer or business estimate. The mistake I saw constantly was candidates trying to recall a specific market statistic they half-remembered, then anchoring their whole tree to a shaky number. Don't. Build up from a figure you're certain about, like total population, and adjust with percentages you can defend. A clean derivation from a known anchor beats a precise-sounding number you can't justify.

Worked example 1 (top-down): coffee cups sold per year in the US

Start with the population: 330 million people. Assume about 75 percent are adults, so roughly 250 million adults. Assume 60 percent of adults drink coffee regularly, giving 150 million coffee drinkers. Now frequency: say the average drinker has 1.5 cups per day. That's 150 million times 1.5 = 225 million cups per day.

Scale to a year. Round 365 days to 360 for clean math: 225 million times 360. Use the zeros system, 225 times 36 = 8,100, then attach the right powers of ten. 225 million is 2.25 times 10 to the 8th; 360 is 3.6 times 10 to the 2nd; the product is about 8.1 times 10 to the 10th, which is roughly 81 billion cups a year.

Now sanity-check. Is 81 billion cups plausible for 150 million drinkers? That's about 540 cups per drinker per year, or roughly 1.5 a day, which is exactly what we assumed, so the loop is consistent. The magnitude (tens of billions) fits a daily-consumption staple in a large country. We'd state the result as "on the order of 80 billion cups a year" and note that the biggest lever is the cups-per-day assumption.

Worked example 2 (bottom-up): annual revenue of one airport coffee kiosk

Now flip directions. Estimate from a single unit. Assume the kiosk is open 16 hours a day. In a busy airport, say it serves an average of 50 customers per hour during operating hours. That's 50 times 16 = 800 customers per day.

Assume the average spend is 6 dollars per customer (a coffee plus a pastry runs higher at an airport). Daily revenue is 800 times 6 = 4,800 dollars. Scale to a year: 4,800 times 360 days. 48 times 36 = 1,728, and with the zeros, 4,800 is 4.8 times 10 to the 3rd, 360 is 3.6 times 10 to the 2nd, product about 1.73 times 10 to the 6th, so roughly 1.7 million dollars a year.

Sanity-check this one too. Is 1.7 million dollars plausible for a single high-traffic airport kiosk? That implies about 4,800 dollars a day, which for a captive-audience location with premium pricing is on the high but believable side. If the interviewer pushes that the kiosk is in a small regional airport, you'd cut foot traffic in half and land near 850,000 dollars, and you'd say so. The point is that your structure flexes cleanly when an assumption changes.

The kinds of prompts you'll actually get

Market-sizing questions arrive in a few recognizable flavors, and knowing which one you're facing tells you whether to go top-down or bottom-up.

When the question is about total consumer demand, top-down from a population almost always wins. When it's about one location, one machine, or one worker's output, bottom-up from a single unit is cleaner and easier to defend. If you're unsure, say which you're choosing and why in one sentence, then go. Interviewers reward a decisive, stated choice far more than a perfect one.

A note on physical-object counts, which trip people up. To estimate traffic lights in a city, don't try to reason about the whole city at once. Estimate the lights in one square mile of a typical urban grid, then scale by the city's area. That representative-slice trick turns an impossible question into a clean bottom-up multiplication, and it generalizes to almost any "how many objects" prompt.

Common traps and how to dodge them

A few errors I saw repeatedly from the interviewer's seat:

• Dropping a zero in the final multiplication. Track powers of ten separately and verify the magnitude against reality before you announce it.
• Over-segmenting. A ten-branch tree is harder to defend than a four-branch one and multiplies your error surface. Keep it shallow.
• Computing in silence. Narrate the tree and each assumption so the interviewer follows and can nudge you back on track.
• Forgetting the "so what." After you land the number, say what it means for the question on the table, since a raw figure with no implication reads like homework, not consulting.

For the underlying arithmetic, the zeros system, friendly-number rounding, and the sense-checks all come straight from the case interview math guide. Market sizing is really just that math applied to estimation.

The bottom line

Market sizing rewards structure and judgment over precision. Pick top-down or bottom-up and say why, build a shallow segmentation tree, anchor every assumption to a known figure and state it aloud, then multiply through and sanity-check the magnitude. A defensible estimate built on a clean tree beats a "right" number you can't explain. Cross-check with the opposite approach when you can.

To see estimation embedded in fuller problems, work through these case interview examples, and to slot market sizing into a complete case flow, read how to structure a case interview.

Go deeper

The Cut to the Case math and structure modules drill market sizing trees, the anchor figures worth memorizing, and the sense-check routines that keep your estimates honest under pressure.

Get the complete Cut to the Case course →

One payment, lifetime access, 30-day guarantee.