TL;DR. On an INSEE base of 68.6M people and 31.4M workforce, a clean Fermi saves you 5 minutes and avoids 80% of silent rejects in round 1. This guide gives you 5 France case studies, the top-down vs bottom-up framework, and the 6 traps that still wreck MBB candidates in 2026.
You have 6 minutes, a marker, and zero numbers in your head. The interviewer drops it on you: how many baguettes are sold per day in France?
Half of candidates rejected in round 1 are rejected on a sizing — not on a strategy case.
What if sizing wasn't a sub-chapter of MECE, but the single most filtering exercise in the whole interview?
Why market sizing is still the most filtering exercise in 2026
Market sizing isn't a maths exercise. It's a test of structure plus numerate common sense. Bain spells it out: "we look for whether you can make sensible assumptions, do quick math, and build constructively on others' ideas" (Bain Hiring Process).
Translation: you're judged on the quality of your assumptions and the cleanness of your tree, not on the fourth decimal.
The anchor stat to memorise: 68.6 million people on January 1, 2025 (INSEE, 2024 Demographic Report). It's the one number you must be able to say without hesitation.
The exercise goes well beyond consulting. A PM pitching TAM/SAM/SOM, a junior VC sizing a pre-seed market, an internal strategist defending a new business case in front of an investment committee: they all use the same Fermi grammar.
Structure first, precision second. That's the rule that separates a ready candidate from one who improvises.
Top-down vs bottom-up: same question, two trains of thought
Bain's CoffeeCo case (Coffee Case Study) asks for the coffee market in Cambridge, residents only. Let's transpose it to Lyon: ~520,000 people intra-muros (INSEE — Lyon municipal dossier).
Top-down: 520,000 × 70% regular drinkers × 1.5 tickets/day × 365 × €2.50 average ticket ≈ €498M/year.
Bottom-up: ~2,000 cafés in Lyon × 200 tickets/day × 365 × €2.50 ≈ €365M/year.
- ✓Population × % consumers × frequency × price
- ✓Lyon coffee result: ~€498M/year
- ✓Fast when demographic anchors are solid
- ✓Risk: double-count if you mix households and individuals
- ✗Outlets × tickets/day × days × average ticket
- ✗Lyon coffee result: ~€365M/year
- ✗More credible on niche B2B or hyper-local retail
- ✗Risk: missing off-channel sales (grocery, vending)
Expected gap: ±20%. Comment on it out loud. Say: "the 25% delta between my two approaches likely comes from off-café sales — supermarkets, office vending. I'd take the average ~€430M for the rest."
That comment is worth as much as the maths. You show you actually read your own numbers.
Calibrating Fermi assumptions with real INSEE 2025 data
Golden rule: always start from a real number. Never "let's say 70 million". The interviewer hears the lazy rounding from the first sentence.
The 7 anchors to memorise before any MBB or PM interview in France:
B2C segmentation: age pyramid (0-14 / 15-64 / 65+), urban/rural, socio-professional category, sometimes gender. For a beauty product targeting women 25-40, start from 68.6M × ~10% in that bracket ≈ 6.9M targets.
B2B segmentation: companies by size (micro / SME / mid-cap / large) and NAF sector. SIRENE lists ~4M active legal units (INSEE — Sirene base); an HR tool for SMEs really targets the ~150,000 companies with 10 to 250 employees (INSEE — Companies in France).
If you say "according to INSEE" out loud during the exercise, you bank a credibility point immediately.
5 France case studies solved step by step
Case 1 — B2C food: daily baguette market
68.6M × 60% daily consumers × 0.8 baguette/day ≈ 33M baguettes/day. Average price €1.10 → ~€36M/day, i.e. ~€13B/year. Sanity check: 0.5% of GDP for an iconic product → coherent.
Case 2 — B2C retail: sneakers, women 25-40
Age bracket ≈ 10% of population (INSEE 2024 Demographic Report) → 6.9M targets × 65% buyers × 1.5 pairs/year × €75 ≈ €504M/year. Cross-check: this segment represents around 15-20% of a France sneakers market in the low single-digit billions → plausible.
Case 3 — B2B SaaS: TAM for an HR tool targeting SMEs
Start from the 403,000 companies using France Travail (France Travail — Companies statistics) as a proxy for hiring-active companies. Filter: ~150,000 SMEs (10-250 employees) (INSEE — Companies in France) × 25% HR SaaS adoption × €8/employee/month × 50 average employees × 12 ≈ €180M/year accessible TAM.
Case 4 — Public sector: training programme entries per year
Anchor: 275,900 entries in Q4 2024 (France Travail — Jobseekers statistics). Annualised: ~1.1M entries/year. Average cost ~€3,500/entry → public training market for jobseekers ≈ €3.9B.
Case 5 — PM tech: TAM for a parenting app
2024 births: 663,000 (INSEE). 0-3 years cohort ≈ 663,000 × 3 ≈ 2M target households × 30% premium smartphone-app adoption × €4.99/month × 12 ≈ €36M annual France TAM.
Key note: every case ends with a sanity check against GDP or an adjacent market. Without that check, a sizing is just an orphan calculation.
The 6 traps that wreck a sizing
- You start from a precise INSEE number (68.6M, not 70M)
- You haven't double-counted (households AND individuals)
- You segmented before multiplying (age, SPC, company size)
- Your time units are consistent (all /year or all /day)
- You handled seasonality where it matters (ice cream, heating oil, toys)
- You close with an explicit sanity check vs GDP or a known market
Trap 1 — Numbers too round. Saying 70M instead of 68.6M costs you credibility in the first minute. You signal to the case lead that you haven't revised.
Trap 2 — Double-counting. Putting 30M households AND 68.6M individuals in the same equation artificially inflates the market 2x.
Trap 3 — Forgotten segmentation. A B2B market without micro/SME/mid-cap/large breakdown gives you a grotesque TAM. Same on B2C without age brackets.
Trap 4 — Time unit. Mixing a monthly volume with an annual price is the #1 junior mistake. Everything in /year, or everything in /day, never hybrid.
Trap 5 — Seasonality ignored. Ice cream in winter, heating oil in summer, toys outside December: weight it or call it out. Otherwise you're off by a factor of 2.
Trap 6 — No sanity check. Always close with: "this number represents X% of GDP (€2,920B) or household consumption (€1,400B), it's coherent." Anchor sources: INSEE — Quarterly national accounts. That single line drives 80% of the gap between an accepted sizing and a rejected one.
Beyond MBB: market sizing for PM tech, VC and internal strategy
Fermi doesn't belong to consulting. It's become the shared grammar of every pre-decision quantitative conversation.
- ✓Goal: filter candidate rigour
- ✓Deliverable: verbal sizing 6-10 min
- ✓Typical duration: 6-10 minutes
- ✗Goal: justify a product bet
- ✗Deliverable: TAM/SAM/SOM pitch slide
- ✗Typical duration: 1-2 hours
- ✓Goal: qualify a pre-seed deal
- ✓Deliverable: 1-pager investment memo
- ✓Typical duration: half a day
- ✗Goal: win an investment committee
- ✗Deliverable: 5-10 slide business case
- ✗Typical duration: 1-2 weeks
On the PM side, the Hacker News community popularised Fermi estimation applied to startup models, and a Wordle-style Show HN on Fermi questions climbed to 35 upvotes (fermiquestions.org). Bain also publishes a second official sizing case in retail (Fashion Case Study) used as a cross-domain reference. Clear signal: the method is now cross-functional.
On the VC side, sizing feeds a 5-year projection with a CAGR; on the internal strategy side, it sits at the base of the investment committee deck before any go/no-go.
Structure over precision. Every time.
FAQ
How much time should I spend on a market sizing?
6 to 10 minutes maximum. Announce your plan in 30 seconds, calculate for 4-5 minutes, close with a 1-minute sanity check. Anything longer and you blow the strategy case that follows.
Top-down or bottom-up: which one?
Top-down for mass B2C markets (food, cosmetics). Bottom-up for niche B2B or supply-constrained services (dental clinics, Michelin-star restaurants). Best practice: announce both and run the faster one.
Do I need the exact French population by heart?
Yes. 68.6 million on January 1, 2025 (INSEE). A candidate who says "around 65 million" in 2026 loses points in the first minute.
How do I handle a sizing question I don't understand?
Rephrase out loud, check the scope (all of France? Paris? B2C or B2B?) and the unit (€ or volume? per year or per month?). 90% of failed sizings come from fuzzy initial framing.
My numbers are wrong — does that disqualify me?
No, if your logic is clean. Bain writes it: "sensible assumptions" matter more than precision (Bain). But broken logic plus wrong numbers equals a near-automatic reject.
Does market sizing show up in PM tech interviews?
Yes, as TAM/SAM/SOM or Fermi questions. The HN community stays active around fermiquestions.org, and Bain also publishes an official retail sizing case (Fashion Case Study).
Which sources can I cite during the exercise?
INSEE for population, workforce and fertility; France Travail (the French unemployment agency) for unemployment, training programmes and companies; Eurostat for EU benchmarks. Citing a real source is worth +1 point.
How do I train efficiently in 2 weeks?
20 timed sizings, one per day, alternating B2C / B2B / public sector. Always close with a sanity check against GDP (€2,920B) or household consumption (€1,400B) — figures from the INSEE quarterly national accounts.
Can sizing target a highly technical B2B market?
Yes, and it's common in internal strategy. Start from the number of companies (INSEE — Sirene base, ~4M active legal units) then filter by NAF sector and size.
Key takeaways
- Sizing is a test of structure and numerate common sense, not mental arithmetic.
- Memorise 7 INSEE 2025 anchors — they cover 80% of the cases.
- Pick top-down OR bottom-up by the 30-second mark, never both in parallel.
- Segment every time (age, SPC, company size) before multiplying.
- Close with an explicit sanity check: "this number is X% of GDP, it checks out".
- The 6 traps (round numbers, double-count, missing segmentation, time unit, seasonality, no sanity check) drive 80% of failures.
- The method is cross-functional: MBB, PM tech, VC, internal strategy — same Fermi, same traps.
Want to drill 20 timed sizings with AI-graded feedback? Check out the Velyq AI interview platform — case interview module with timed sizing and the MBB grid built in.
Does your CV pass the consulting filter? Run a free CV review before your next MBB process.
