A guesstimate is a Fermi estimate: break an unknown into factors you can bound, keep units, and land an order of magnitude a skeptical partner could live with. It is not a market-sizing TAM with a price. If nobody is buying, you are not sizing a market. You are counting.
Interviewers use these to see whether you invent structure under uncertainty — not whether you memorized “US population 320 million.”
Guesstimate versus market size
| Guesstimate | Market size | |
|---|---|---|
| Question shape | How many / how much physical stuff | How many $ of spend |
| Output unit | planes, beds, trips, tonnes | $ / year |
| Price | Usually absent | Required |
| Kill test | Order of magnitude vs a known capacity | Bottom-up vs top-down spend |
“How many MRI scans in Cuyahoga County last year?” is a guesstimate. “US spend on outsourced warehouse HVAC service” is a market size. Do not force a $/year on the first.
Worked example: Wednesday takeoffs at a mid-size airport
Prompt. How many commercial passenger takeoffs does Raleigh–Durham (RDU) have on a typical Wednesday? (Not passengers. Not cargo-only. Takeoffs.)
Clarify. Commercial = scheduled passenger airlines. Wednesday = midweek, not Sunday. I will estimate departures, not movements (takeoff + landing would double-count).
Structure (supply-side, because airports have visible capacity).
Takeoffs / day ≈ gates in use × turns per gate per day × % of turns that are departures from this count (each turn is one arrival + one departure, so departures ≈ turns if we define a turn as a departure).
Cleaner: departures / day ≈ aircraft departures = (gates) × (departures per gate per day).
Gates. A mid-size connecting-ish airport: RDU is not ATL. Reasonable band 40–50 contact gates. Use 45 gates. (If I am wildly off, the interviewer will nudge. I will also sanity-check with flights/hour.)
Turns. Domestic banks: morning out, midday, evening. A gate that is busy might turn 6–8 times a day; a spoke gate 3–4. Blend: 5 departures per gate per day.
45 × 5 = 225 departures / day.
Cargo, GA, international thinning. Commercial passenger only: knock 10% for gates used by oversized or idle. ~200.
Hourly sanity check. Operating day ~18 hours (6am–midnight). 200 / 18 ≈ 11 departures per hour. Peak banks maybe 2× that = 22/hour. One runway can handle ~40–50 operations/hour including landings; 22 takeoffs + 22 landings = 44 operations. Physically fine. 800/day would not be: that is a hub.
Wednesday vs Friday. Wednesday is ~90% of a Friday. Negligible for order of magnitude. Keep ~200.
Range. Gates 40–50 (±11%). Turns 4–6 (±20%). Combined roughly 150–270. I would present about 200 commercial passenger takeoffs on a Wednesday, plus/minus a third.
If the interviewer says RDU has ~35 gates, update: 35 × 5 = 175. Same order.
How to run any guesstimate
- Restate the unit. Takeoffs ≠ passengers ≠ seats.
- Pick supply or demand, whichever has a tighter capacity constraint. Airports, hospitals, and kilns are supply-shaped. Consumer habits are demand-shaped.
- Three to five factors, not twelve. Extra factors fake precision.
- Bound one factor you are least sure of (“gates 40–50”).
- Sanity check against a known ceiling (runway rate, beds × occupancy, nameplate tonnes).
Write the chain:
takeoffs/day = gates × departures per gate per day
not a paragraph of vibes.
When a guesstimate is the wrong tool
The case already has an exhibit with the number. Do not Fermi a figure they handed you.
They asked for a market in dollars. Switch to step-by-step sizing and put price in the chain.
They asked why profit fell. Counting planes will not isolate a driver. Use profitability.
The mistake unique to guesstimates
Counting the wrong object. Candidates estimate passengers when asked for takeoffs, then multiply by nothing. 200 takeoffs × 150 seats × 85% load is ~25,000 passengers — a useful check, but it is a different question. Label the unit every line.
The other unique fail is starting from US population for a local capacity object. RDU takeoffs do not start at 330 million people. They start at gates and banks. Population is for demand-side consumer Fermi problems (“piano tuners”), not infrastructure.
What “good” sounds like
“I will estimate commercial passenger departures, not passengers. Mid-size airport, about 45 gates, about five departures a gate, two hundred a day. Runway math still works. If gates are thirty-five, I land near one seventy-five. Order of magnitude is low hundreds, not thousands.”
That is a partner-safe Fermi. A single fake-precise 187 is not.
If they follow with “now passengers,” do not restart from population. Reuse the 200 takeoffs: 200 × 140 seats × 82% load ≈ 23,000 passengers / day. You kept the unit chain. That is how a guesstimate turns into a later market size if they add a ticket price. Until they add $, you still do not have a TAM.
Count with units, then check a ceiling
Market-sizing drills also train Fermi chains. Use them for how-many questions, not only TAM.