The Four Approaches
Two axes, four paths — and the one move that separates a good answer from a defensible one.
Two candidates get the same prompt — how many air conditioners are sold in India each year? The first starts listing every AC brand she can name. The second says: I'll go top-down from households, then sanity-check from the supply side. She hasn't computed anything yet, and she's already ahead — because the path you pick in the first fifteen seconds decides whether the next three minutes are a clean walk or a scramble. There are only four paths. Knowing all four — and which one fits the prompt — is the whole game of the Set-Up stage.
TL;DR · Key Takeaways
After this page, you will be able to
- Place any guesstimate on two axes — Direction (Top-Down vs Bottom-Up) and Lens (Demand vs Supply) — and name its quadrant out loud.
- Pick the path by where your anchors are strongest, not by habit, and recognise when too many filters will compound your error.
- Reach for the supply side when everyone else reaches for demand — often the faster anchor and an instant cross-check.
- Solve in one quadrant and re-derive from the diagonal, treating a large gap as a diagnostic rather than something to average away.
- Map each of the three flavours to its default path so your Set-Up is decided in the first fifteen seconds.
Two axes, not four random tricks
People talk about "approaches" as a grab-bag of tricks. They're not. Every guesstimate path is a choice on two independent axes. Pick a point on each axis and you've picked your approach — and, just as usefully, you've named the other approach you'll triangulate against later.
| Axis | One end | Other end | The question it answers |
|---|---|---|---|
| Direction | Top-Down | Bottom-Up | Do I start from a big universe and filter down, or from one unit and scale up? |
| Lens | Demand-side | Supply-side | Do I count the people who want the thing, or the people/assets that provide it? |
Direction is about where you start. Lens is about what you count. They combine into a 2×2 — four quadrants, four defensible paths to the same number. The casebooks call the top-down move "driver breakdown" and the bottom-up move "extrapolation"; the names matter less than knowing both directions exist and that a strong solve usually walks one and checks the other.
When each path wins
You don't pick a quadrant by taste — you pick it by where your anchors are strongest. An anchor is a number you can defend without flinching: India's population, a household's monthly grocery bill, the seats in a multiplex. Start where your anchors are solid; the path that forces you to invent a number you have no feel for is the wrong path.
You have a clean universe
You can picture one unit clearly
Output is easier to anchor than users
"Smartphones in India" — start from population, filter by age, income, ownership.
"Revenue of a Bandra café" — size one café's day, then don't even need to scale.
"Petrol pumps" — national fuel demand ÷ per-pump throughput beats counting drivers.
Risk: too many filters compound error.
Risk: your one unit isn't representative.
Risk: forgetting utilisation (assets sit idle).
The flavour usually names the path
Recall the three flavours from Page 1. Market Sizing prompts lean Top-Down + Demand (filter a population). Revenue / Throughput prompts lean Bottom-Up (size one unit per period). Supply-side is the contrarian's edge: when everyone reaches for demand, counting the providers often gives a cleaner, faster anchor — and instant triangulation.
The move that separates the answers
Here is the single highest-leverage habit on this page: solve in one quadrant, then re-derive from the diagonal. Top-Down Demand checked by Bottom-Up Supply. If the two land within ~25% of each other, you state the number with confidence. If they're 5× apart, you've found your own error before the interviewer did — which scores higher than a single lucky-perfect number ever could.
Same prompt, two quadrants — and the reconciliation
How many petrol pumps are there in India?
I'll go Top-Down from the supply side — national petrol demand divided by what one pump moves — and then cross-check Bottom-Up from vehicles. Let me start.
Names both the path and the cross-check upfront — a Set-Up scoring move before any number.
Petrol demand is roughly 40 billion litres a year. A busy outlet does maybe 150,000 litres a month, so ~1.8 million a year. 40B ÷ 1.8M ≈ ~22,000 — but many rural outlets do a fraction of that, so blend down the average and it's closer to ~80,000 outlets.
Bottom-up sanity check: ~300 million vehicles, of which maybe a third are petrol two-wheelers and cars filling, say, weekly. If one pump can serve ~1,500 vehicles as regulars, you land in the same ~70–90k band. Two methods agree — I'll call it ~80,000, with the per-pump throughput as the swing assumption.
The diagonal check. Two independent builds, reconciled out loud.
The number was never the point. The score came from naming two paths, walking one, and reconciling the other — the exact discipline the rest of this section drills.
Don't average two methods you don't believe
Triangulation reconciles, it doesn't blend blindly. If Top-Down says 80k and Bottom-Up says 400k, the answer is not 240k — it's "one of my assumptions is wrong, let me find which." The gap is a diagnostic, not an arithmetic mean. Page 4 turns this into a repeatable pressure-test.
Where this connects
This page expanded the Set Up stage of the spine. Page 3 — The Ideal Flow trains the stages on either side of it: how to Clarify the prompt so you pick the right quadrant, and how to communicate the solve while you walk it. Page 4 — Pressure-Testing makes the diagonal cross-check and the sanity check non-negotiable. The India cheat sheet (Page 5) gives you the anchors these approaches lean on. Every worked example in this section is tagged with the quadrant it uses — read three and you'll start seeing the path before the prompt finishes.