The Pain & The Promise
Why most candidates lose this case before they touch a number.
Three minutes in, the candidate confidently delivers her number: ₹4,000 cr a year for premium dark chocolate in India. The interviewer leans back. India's entire chocolate market is around ₹15,000 cr. You're saying premium dark alone is a quarter of that? She freezes. The structure was right. The math was right. The order of magnitude was unforgivable. Guesstimates don't test arithmetic — they test whether you can be trusted with a ₹40 cr decision without a spreadsheet. And the candidates who can are not the ones who memorised India's population.
TL;DR · Key Takeaways
After this page, you will be able to
- Spot what the grader is actually scoring — and recognise that 75% of the score lives outside arithmetic.
- Name a guesstimate's flavour out loud (Market Sizing / Revenue / Unconventional) so your default approach is set before you start.
- Move through the five-stage spine in order — Clarify → Set Up → Estimate → Triangulate → Recommend — without skipping a stop.
- Triangulate every solve with a second method, even a rough one — skipping it is the single biggest source of unforced point loss.
- Recognise when a prompt blends flavours and solve the layers separately instead of forcing one bucket.
What's actually being graded
Most candidates train guesstimates the way they trained for board exams — drilling mental arithmetic. Then they walk into a case where the math is rounded to the nearest crore and a wrong-by-10× answer ends the round in 90 seconds. The grader's rubric tells a different story: arithmetic is 15% of the score. The other 85% is what to compute, how to split it, and whether you noticed when you were obviously wrong.
| Dimension | Weight | What it tests |
|---|---|---|
| Structure & Approach | 30% | Did you write the equation BEFORE you started multiplying? |
| Segmentation & Assumptions | 25% | Is your split MECE? Is each number justified out loud? |
| Sanity Check & Triangulation | 20% | Did you cross-check with a second method? |
| Arithmetic & Unit Discipline | 15% | Right math, right units, no drift between steps. |
| Scoping & Clarification | 10% | Did you lock the prompt before you solved it? |
Two-thirds of your score sits in moves that happen before you do any multiplication — picking the equation, picking the splits, choosing whether to triangulate. A candidate who lands a perfect arithmetic score but skips Triangulate forfeits twenty points she could have caught in fifteen seconds. This is the leverage.
Three flavours, one skill
Interviewers ask guesstimates in three flavours, and the flavour you're handed sets your default approach. Mistake a Revenue prompt for a Sizing prompt and you'll burn ninety seconds estimating the wrong universe. Name the flavour out loud and the rest of the solve clicks into a known shape.
How big is this market?
What does this do per day or year?
Estimate the unestimatable.
Demand for ACs in India this year
Daily revenue of a Bandra café
Tennis balls that fit in an IPL stadium
Top-down → Demand
Bottom-up → Supply or Demand
Driver breakdown under ambiguity
₹ value × unit volume
₹ per period
A defensible order of magnitude
Not every prompt fits cleanly. Revenue of food-delivery apps in Delhi over the next five years stacks sizing on throughput on growth. The discipline is to spot the stack and solve the layers separately — not to find a single perfect bucket and force-fit the prompt into it.
The 90-second spine
Strip away the casebook noise and every guesstimate you'll ever solve travels the same five stops. The candidate who delivers the strongest answer is rarely the one with the cleverest split — she's the one who visited all five stops in order and stayed at each long enough to do the work.
The skip that ends most cases
Triangulate is the cheapest move on the page — a second method, even a quick supply-side check, takes fifteen seconds — and the move most candidates skip. The chocolate failure at the top of this page would have been caught by any cross-check. Make Triangulate a non-negotiable, not a flourish.
The same prompt, twice — first without the spine, then with it
Estimate the annual demand for premium dark chocolate in India, in rupees.
OK — India is 1.4 billion people, urban India is ~30%, so 42 cr urban. Middle class is, say, 40% of urban — 17 cr. Premium-affording is 20% of that — 3.4 cr. Half eat chocolate at all, so 1.7 cr. Maybe 30% prefer dark — 50 lakh. At 6 bars a year times ₹400… wait, let me redo… 50 lakh times ₹2,400… about ₹1,200 cr. Actually let me re-check the urban share… about ₹4,000 cr per year.
Layered filters with no equation written first; numbers shift mid-stream and the candidate rebuilds the calc twice.
India's entire chocolate market — Dairy Milk, Munch, KitKat, the lot — is roughly ₹15,000 cr. You're saying premium dark alone is a quarter of that?
…I'd have to revisit my assumptions.
No second method, no anchor, no recovery path. This is where the case is lost.
Same prompt, with the spine.
Before I solve, let me lock it: premium dark chocolate, India, annual, in rupees. I'll work top-down — estimate buyers, then average spend — and sanity-check from the supply side at the end.
Fifteen seconds to clarify the prompt, name the approach, and flag the cross-check upfront. Three scoring moves before a single number is touched.
Urban India: ~50 cr. Top-10% premium-discretionary households: 5 cr people. Of those, dark-chocolate consumers — maybe 1 in 5 — so 1 cr active buyers. At 4 bars a year × ₹250: ₹1,000 cr. Supply side: Lindt, Amul Dark, Bournville premium SKUs, Hershey's, imports — about six serious players. If the leader does ₹200 cr in India, total is plausibly 4× — ₹800 cr. Two methods within 25%. Central estimate ₹900 cr.
Sensitivity?
The active-buyer ratio swings it. 1 in 3 instead of 1 in 5: ₹1,700 cr. 1 in 10: ₹500 cr. Range ₹500–1,700 cr, central ₹900 cr.
Same prompt. Same candidate. The redo scored on Structure (equation first), Triangulation (two methods reconciled), and Bounds (sensitivity explicit). Pages 2–5 train each of those moves.
Where this connects
This page is the on-ramp. Page 2 unpacks the four approaches (Top-Down × Bottom-Up × Supply × Demand) and how to pick between them. Page 3 trains the Ideal Flow — clarify, set up, communicate — with the live scripts. Page 4 is Pressure-Testing — triangulation, magnitude checks, common traps. Page 5 is the India cheat sheet — the numbers worth memorising and how to deploy them live. Page 6 is one full solve, annotated. Outside this section: Pricing teaches willingness-to-pay estimation, TAM / SAM / SOM is the formal nest for sizing, and the 5 C's is the upstream lens for who you're sizing for.