Interviewers rarely circle the final dollar figure. They circle the chain: whether you defined the object, put assumptions on the table, and ran a second path that lands in the same order of magnitude. A precise $291.4 billion with one brittle funnel loses to a $250–350 billion band with a visible challenge path.
That is the thesis of this page. Market sizing is scored as structured uncertainty, not as a prediction contest. After you finish, you should be able to size one market two ways, reconcile them aloud, and know which cell the interviewer will poke.
For question types and what the prompt is testing, see market sizing questions. For Fermi counts without a price, see guesstimates.
What the scoring sheet actually rewards
In a typical first round of eight candidates, two or three produce a clean size. The rest either invent a US-population funnel that never fits the object, defend a fake-precise number, or freeze when one assumption is challenged.
| What you do | How it usually lands |
|---|---|
| Define the object in one sentence the interviewer repeats back | Marks for scope control |
| Write 4–7 factors with units that cancel to $/year (or units/year) | Marks for structure |
| Table base / low / high so a challenge hits one cell | Marks for coachability |
| Run a second path (demand vs supply, or top-down vs bottom-up) | Marks for sanity |
| State the band and the so-what for the decision | Marks for judgment |
| Present six significant figures as "the answer" | Soft fail, even if arithmetic is right |
I stop listening when the candidate treats the base case as sacred. The table is the answer.
Define the object before you touch a number
Prompt we will size (original, not a firm case): annual US consumer spend at limited-service restaurants — QSR and fast casual, dine-in and takeaway, excluding full-service restaurants, grocery prepared food, and meal-kit delivery.
Say that sentence out loud. If the interviewer wants specialty coffee only, or wants to include grocery deli, you adjust before you multiply. Most sizing failures start here: people × % × $ with no agreed object.
Units for this page: dollars per year, US, consumer spend (what households pay), not franchise royalty pools.
Path 1 — top-down from BLS food away from home
Start from a published spend bucket, then slice.
As of the BLS Consumer Expenditure Surveys for 2023, the average consumer unit spent $3,933 on food away from home, and there were 134.556 million consumer units.
3,933 × 134.556 million ≈ $529 billion.
That $529B is all food away from home: full-service, limited-service, workplace cafeterias, and related away-from-home meals. Limited-service is the majority of occasions but not all of the dollars (tickets are lower). For interview purposes I use a 55% limited-service share of FAFH dollars — labeled as Ned's estimate, not a BLS line item.
$529B × 0.55 ≈ $291 billion.
Sensitivity if the share is 45–65%:
| Limited-service share of FAFH | Implied US limited-service spend |
|---|---|
| 45% | ~$238B |
| 55% (base) | ~$291B |
| 65% | ~$344B |
Top-down answer so far: roughly $240–340 billion per year, base near $290B.
Path 2 — bottom-up from Census population
Same object, different chain. Anchor on population from the Census Bureau's December 2024 population estimates release: 340.1 million US residents as of July 1, 2024. For household-level products you would start from ACS households instead; for QSR occasions, population is the cleaner stock.
| Factor | Base | Why |
|---|---|---|
| US population | 340M | Census July 1, 2024 estimate |
| Share who buy limited-service at least weekly (incl. parents buying for kids) | 55% | Interview assumption — challenge freely |
| Visits per buyer per week | 2.5 | Mix of daily coffee/QSR users and weekly families |
| Weeks per year | 52 | Calendar |
| Average ticket (pre-tip, incl. combo) | $12 | Interview assumption |
Arithmetic:
340M × 0.55 = 187M regular buyers.
187M × 2.5 × 52 = 24.3 billion visits per year.
24.3B × $12 = $292 billion.
Same order as the top-down base ($291B). That is not luck if both chains share a realistic intensity; it is the check working.
If bottom-up had landed at $2.9 trillion, you would kill the visit rate or the ticket before you ever said the number out loud. If it landed at $29B, you would check whether you sized specialty coffee instead of all limited-service.
Reconcile, then give the band
| Path | Base | Implied band |
|---|---|---|
| Top-down (BLS FAFH × limited-service share) | ~$291B | $240–340B |
| Bottom-up (pop × buyers × visits × ticket) | ~$292B | $200–400B if you move visits or ticket ±25% |
| Reconciled answer | ~$250–350B | Order-of-magnitude: hundreds of billions |
Do not average 291 and 292 into fake precision. Say: about three hundred billion dollars a year, give or take a third. For a "should we buy a regional QSR platform" decision, $0.3T vs $30B vs $3T is what matters. We are in the hundreds of billions.
Cross-check with other official populations when the case is international. Eurostat put the EU population at 449.2 million on 1 January 2024. ONS mid-2023 estimates put the UK at 68.3 million. Those are anchors for a parallel size abroad — not multipliers you apply blindly to US spend intensity without saying so.
Ned's rule. Score yourself on whether a second path lands in the same power of ten. If it does not, fix a factor before you "answer." The interviewer is grading the fix, not the first draft.
Assumptions table (challenge one cell)
| Assumption | Base | Low | High | Effect on $292B base |
|---|---|---|---|---|
| Weekly buyer share | 55% | 40% | 70% | Linear |
| Visits / buyer / week | 2.5 | 1.5 | 3.5 | Linear |
| Average ticket | $12 | $9 | $15 | Linear |
| Limited-service share of FAFH (top-down) | 55% | 45% | 65% | Linear on $529B |
| Implied TAM | ~$290B | ~$150B | ~$500B | Still "hundreds of billions" |
When the interviewer says "make visits 1.5," you do not rebuild from population. You scale: $292B × (1.5 / 2.5) ≈ $175B, then re-check whether top-down still fits or whether your FAFH share was too high.
How interviewers listen while you write
Talk the structure first:
"I will size annual US limited-service restaurant spend in dollars per year. Top-down from BLS food-away-from-home times a limited-service share. Bottom-up from Census population times weekly buyers times visits times ticket. I will table assumptions and reconcile to a band."
Then fill the table. Then the band. Then the so-what: a ~$300B pond is a national platform market, not a niche specialty — entry is share fight and unit economics, not "is there a market."
What leaks points in the room:
- A chain that does not cancel to $/year (people × % with no price, or visits with no ticket).
- Silent borrowing of "300 million Americans" when you needed households, establishments, or consumer units.
- Refusing to move one cell when challenged.
- Skipping the second path because "I already have a number."
When not to run a full size
The exhibit already gave you the TAM. Use it and move to share, growth, or unit economics.
The decision is a 2% price cut or a store-level margin fix. You need contribution and elasticity, not national spend. Size is theatre.
You cannot define the object in one sentence. Stop. "Restaurants" without limited vs full service is two markets.
You are in a chart-heavy case and the clock is dying. A 30-second order-of-magnitude ("tens of billions, not hundreds") beats a seven-factor model you cannot finish.
Practice this today
Close this tab. Time yourself for 12 minutes on a new object — not QSR. Write the object sentence, a four-to-seven factor chain, an assumptions table, and a second path. Then check your arithmetic on the market size calculator and drill more prompts on market-sizing drills.
Everything on CoachNed is open for seven days, no card; then $120 for a recruiting season or $49 a month. Pair sizing with case math practice so unit errors do not masquerade as bad judgment.
CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any firm named in interview folklore.
Frequently asked questions
How long should a market sizing take in a case interview?
Three to eight minutes for a clean size inside a broader case; up to about twelve if the prompt is the sizing. Structure and a second path beat extra factors.
Do I need the exact US population memorized?
You need a defensible anchor within about 10%. As of the Census July 1, 2024 estimate, use roughly 340 million. Say the source vintage if you know it; interviewers care more that you flag uncertainty than that you hit the latest release.
Top-down or bottom-up first?
Either. Pick the path with the better published anchor for this object, then force the other as the check. Food and consumer goods often start from BLS or household counts; B2B often starts from establishments or end customers.
What if my two paths disagree by 10x?
Treat it as a bug, not a debate. Revisit the largest soft factor (penetration, frequency, price, or share of a published bucket) until they share an order of magnitude, then report the band.
Is the final number what gets me the point?
No. The point is the defined object, the assumption table, the reconciliation, and the so-what for the decision. A wrong-but-coherent band outperforms a lucky exact figure you cannot defend.
Sources
- Consumer expenditures in 2023 (BLS) — consumer units (134.556 million) and food away from home ($3,933) for the top-down path. Checked 2026-09-24.
- New 2024 population estimates (U.S. Census Bureau) — US population 340.1 million as of July 1, 2024 for the bottom-up path. Checked 2026-09-24.
- EU population increases again in 2024 (Eurostat) — EU population 449.2 million on 1 January 2024 for international cross-checks. Checked 2026-09-24.
- Population estimates for the UK: mid-2023 (ONS) — UK population 68.3 million at mid-2023 for scaled international checks. Checked 2026-09-24.
- Consumer Expenditures — 2023 news release (BLS, archived) — corroborating consumer-unit and FAFH figures from the official release. Checked 2026-09-24.