I stop writing when a candidate delivers a confident 187,432 and cannot say which factor would move the answer by half. On the sheet, the number is almost never the point. The chain is.
Thesis: a guesstimate is scored on structure, units, and a ceiling check a partner could live with — not on whether you secretly knew the true figure. Dollar market sizing is a different drill; that sibling lives in Market sizing step by step. This piece is the worked question bank: six original Fermi chains across population, supply, replacement, and time-and-motion structures, each anchored to a public statistic, then the five ways people leak marks.
After you finish, you should be able to pick a structure in under thirty seconds, write a three-to-five-factor equation, and sanity-check the order of magnitude before the interviewer asks.
What the sheet actually rewards
| What you do | How it lands |
|---|---|
| Restate the unit out loud | Stops the wrong-object fail |
| Write 3–5 factors, not twelve | Looks like judgment, not theatre |
| Bound the weakest factor | Shows you know where uncertainty lives |
| Run one ceiling or floor check | Separates partner-safe from fake-precise |
| Update cleanly when nudged | Recovers points after a bad assumption |
In a typical first round of eight candidates, two or three produce a clean chain. Most of the rest either start from population for a capacity object, or invent a dollar TAM nobody asked for.
Ned's rule. If you cannot write
answer = A × B × Con one line before you multiply, you are not estimating — you are narrating. Force the equation first; the arithmetic is the easy part.
Six worked guesstimates
Each prompt below is original. Each chain starts from a public anchor you can cite in the room if asked ("Census puts the US near 340 million"). The final figure is a band, not a fake integer.
1. Population-based — haircuts on a typical Tuesday (US)
Prompt. How many haircuts occur in the United States on a typical Tuesday?
Anchor. The Census Bureau put the US resident population at about 341.8 million as of July 1, 2025.
Clarify. Paid cuts at salons and barbers only. Tuesday = midweek, not Saturday peak. Exclude home cuts.
Chain.
haircuts/Tuesday ≈ population × share who get paid cuts × cuts per person per year × (1/52) × Tuesday share of weekly cuts
- Population: 340 million (round the Census figure; order of magnitude is what matters).
- Share with paid cuts at least once a year: 70% (kids, home cuts, and bald heads knock the rest).
- Cuts per person per year among that group: 8 (roughly every six weeks).
- Spread evenly across weeks:
340M × 0.7 × 8 / 52 ≈ 37 million/week. - Tuesday is quieter than weekend: take 12% of weekly volume (weekday share of a seven-day week, nudged down).
Result. 37M × 0.12 ≈ 4.4 million haircuts on a Tuesday. Band: 3–6 million.
Sanity. 4.4M / 340M ≈ 1.3% of the country gets a cut that day — one person in seventy-five. Feels right for a midweek service day. Ten million would imply one in thirty-four; that is Saturday talk.
2. Supply-side — occupied US hospital beds tonight
Prompt. How many staffed hospital beds in the United States are occupied right now (a typical non-surge night)?
Anchor. The American Hospital Association reports 907,216 staffed beds across US hospitals (2024 AHA Annual Survey, Fast Facts 2026).
Clarify. Staffed inpatient beds, not licensed capacity or ED stretchers. Night = overnight census, not ambulatory chairs.
Chain.
occupied beds ≈ staffed beds × average occupancy
- Staffed beds: 900,000.
- Occupancy on a normal night: 65–70% (pre-surge community hospitals often sit in this band; I use 68%).
Result. 900,000 × 0.68 ≈ 610,000 occupied beds. Band: 550,000–650,000.
Sanity. AHA also reports about 35.7 million admissions per year. At a 4.5-day average length of stay, bed-days ≈ 35.7M × 4.5 ≈ 160M. Divided by 365 ≈ 440,000 average occupied — same order as 610,000 once you remember staffing includes psych and specialty beds with different utilization. Order of magnitude is hundreds of thousands, not millions.
3. Replacement-rate — light-vehicle tires replaced per year (US)
Prompt. How many passenger and light-truck tires are replaced in the United States in a typical year (excluding brand-new vehicles' original equipment)?
Anchor. FHWA Highway Statistics 2024 put total registered motor vehicles at 297.5 million.
Clarify. Replacement tires only. Drop motorcycles (~9 million in the same table). New-car OE tires are out of scope.
Chain.
tires/year ≈ (registered light vehicles) × (tires per vehicle) × (1 / life in years)
- Light vehicles (all motor vehicles minus motorcycles): ~290 million.
- Tires per vehicle in the replacement pool: 4.
- Average life before replacement: 4 years (mix of commuting and highway; performance tires shorter, trucks longer).
Result. 290M × 4 / 4 = 290 million tires per year. Band: 200–350 million.
Sanity. That is roughly one tire per registered vehicle per year. If life were eight years, you would land near 145 million — still hundreds of millions, not billions. Spares and seasonal swaps sit inside the band.
4. Time-and-motion — full-time piano tuners in the US
Prompt. How many people in the United States earn a living primarily as piano tuners?
Anchor. Same Census population: about 342 million. Census Historical Households Tables (CPS) put recent US household totals near 130 million; I use that round figure.
Clarify. Full-time equivalent tuners, not hobbyists. Acoustic pianos only.
Chain.
tuners ≈ (households × piano ownership × tunings per piano per year) / (tunings one tuner completes per year)
- Households: 130 million.
- Share with an acoustic piano: 2% (Ned's estimate; rarer than people think).
- Tunings per piano per year: 1.
- Tunings per tuner per year: 4 per day × 250 workdays = 1,000.
Result. (130M × 0.02 × 1) / 1,000 = 2,600 tuners. Band: 1,500–5,000.
Sanity. If ownership were 5%, you get ~6,500 — still low thousands. A six-figure answer would require almost every household to own a piano. Time-and-motion is the constraint: the bottleneck is how many tunings one person can finish, not how many pianos exist.
5. Population / household — washing-machine loads on a weekday
Prompt. How many residential washing-machine loads run in the United States on a typical weekday?
Anchor. Roughly 130 million US households from Census CPS household tables.
Clarify. In-home machines only. Exclude laundromats. Weekday = Monday–Friday average.
Chain.
loads/weekday ≈ households × share with a washer × loads per household per week × (weekday share)
- Households: 130 million.
- Share with an in-unit or in-building washer: 80%.
- Loads per household per week: 5.
- Weekday share of weekly laundry: 5/7 ≈ 70% (weekends catch some bulk loads; I use 65% of weekly volume across five weekdays, so each weekday gets
0.65/5 = 13%of weekly — cleaner: weekly loads × 0.65 / 5).
Weekly loads: 130M × 0.8 × 5 = 520 million/week.
Weekday: 520M × 0.65 / 5 ≈ 68 million.
Result. About 70 million loads on a weekday. Band: 50–90 million.
Sanity. EIA puts average US household electricity use near 10,500 kWh per year. A washer is a small slice of that bill; 70 million loads does not break the residential power budget. One load per household every other day among washer owners is the right feel.
6. Stock-from-flow (replacement cousin) — disposable diapers per day
Prompt. How many disposable diapers are used in the United States on a typical day?
Anchor. CDC/NCHS final data: 3,628,934 US births in 2024.
Clarify. Disposable only. Newborn through toddler use, not adult incontinence.
Chain.
diapers/day ≈ births/year × years in diapers × diapers per baby per day
- Births/year: 3.6 million.
- Years in disposables: 2.5.
- Diapers per baby per day: 6.
Stock of babies in diapers: 3.6M × 2.5 ≈ 9 million.
Daily use: 9M × 6 = 54 million.
Result. About 50–60 million diapers per day. Band: 40–80 million.
Sanity. Annual volume ≈ 54M × 365 ≈ 20 billion. At a few cents of manufacturing cost per unit, that is a multi-billion-dollar category — which matches how large the diaper aisle feels without requiring a price. Wrong object would be counting packs, not diapers.
Five ways candidates lose points
| Leak | What it sounds like | Fix |
|---|---|---|
| Wrong object | Estimates passengers when asked for takeoffs | Restate the unit before factor one |
| Wrong structure | Starts from national population for a gate-constrained airport | Ask: is capacity or demand the tighter bound? |
| Fake precision | "187,432" with no band | Give a mid-point and a ±30–50% range |
| No ceiling | Lands at a figure that exceeds a known physical limit | Check beds, runway rate, hours in a day, or stock |
| Silent update | Freezes when the interviewer corrects a factor | Recompute out loud in ten seconds |
The unique guesstimate fail, separate from market sizing, is counting the wrong physical object and then refusing to relabel. Market-sizing fails are usually missing price. Do not confuse the two.
How to pick a structure in thirty seconds
| Clue in the prompt | Default structure |
|---|---|
| "How many people / households / trips…" with no hard capacity | Population or household |
| Hospitals, airports, factories, fleets, beds, gates | Supply-side |
| "Replaced," "worn out," "sold as replacement," "tires," "phones" | Replacement-rate (stock / life) |
| "Tuners," "baristas," "inspectors," any craft with a clear cycle time | Time-and-motion |
| "Babies using X," "cars needing Y each year" from a flow | Stock-from-flow |
If they later ask for dollars, hand the count to a market-sizing chain and add price. Until then, leave $ out of the equation.
Practice this today
Open one prompt from the six above. Write the equation before any arithmetic. Bound one factor. Run one sanity check. Then do three more timed reps on CoachNed's market-sizing drills — the same muscle, without forcing a TAM. Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
For the mental-math hygiene that sits under every Fermi (units first, then multiply), use Mental math in case interviews and the free quick-math drill.
CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any firm named here.
Frequently asked questions
What is a guesstimate in a consulting interview?
A guesstimate is a Fermi estimate: break an unknown quantity into a few factors you can bound, keep units, and land an order of magnitude. It answers "how many" or "how much stuff," not "how many dollars of spend."
How is a guesstimate different from market sizing?
Market sizing multiplies a volume by a price to get annual spend. A guesstimate usually stops at the physical count. If nobody is buying, you are not sizing a market — you are counting. See the sibling guide Market sizing step by step when the prompt asks for dollars.
Do I need to memorize population and other statistics?
Memorize a handful of anchors (US population ~340 million, households ~130 million, roughly 160 million employed). Everything else is structure. Interviewers care more that you state an assumption and check it than that you recalled a precise Census release.
How accurate does my final number need to be?
Within a factor of two or three of a defensible mid-point is usually enough, if the chain is clean. A wrong structure with a lucky number still fails. A clean chain that updates when nudged usually passes.
Should I use top-down or bottom-up?
Use whichever has the tighter constraint. Capacity objects (beds, gates, machines) want supply-side. Consumer habits want population or household. Replacement questions want stock divided by life. Do not force population into every prompt.
Sources
- U.S. population growth slows (Census Bureau, Vintage 2025) — US resident population ~341.8 million as of July 1, 2025. Checked 2026-09-24.
- Historical Households Tables (U.S. Census Bureau) — CPS household totals near 130 million used as the household anchor. Checked 2026-09-24.
- Fast Facts on U.S. Hospitals, 2026 (American Hospital Association) — staffed beds (~907,216) and admissions. Checked 2026-09-24.
- FHWA Highway Statistics 2024, Table MV-1 — registered motor vehicles (~297.5 million). Checked 2026-09-24.
- Births in the United States, 2024 (CDC/NCHS Data Brief 535) — 3,628,934 US births in 2024. Checked 2026-09-24.
- Electricity use in homes (U.S. EIA) — ~10,500 kWh average household electricity use per year. Checked 2026-09-24.