Guesstimate interview practice with a Noida metro UPI example

The 12-minute plan

Write the tree on paper or the whiteboard as you go. Interviewers follow the structure far more easily than a spoken stream of numbers.

Stage 1: clarify before multiplying anything

Prompt: "Estimate the number of UPI transactions made in a day at Noida Sector 18 metro station."

Good clarifying questions, and the answers you might assume if the interviewer says "your call":

• Inside the station only, or around it? Assume merchants inside the station and within about a five-minute walk of its exits. Sector 18 station on the Blue Line sits beside a large shopping market, so "around it" is where most payments will be.
• Which day? Assume a typical weekday, and handle the weekend at the end.
• Metro ticket and card top-ups paid by UPI? Include them as a separate, small line.
• Person-to-person transfers? Exclude them; count payments to merchants.

Stating the scope in one sentence at the end of this stage — "weekday UPI payments to merchants within five minutes' walk of the station, plus ticketing" — is worth marks by itself.

Stage 2 and 3: the demand-side estimate

Break it down by people who pass through and buy something.

1. 1People in the area during the day. Commuters using the station plus shoppers and office workers nearby. Assume 60,000 distinct people on a weekday. Reason: a busy station beside a shopping market, with many people passing twice (in and out) but counted once.
2. 2Share who buy anything nearby. Many commuters walk straight through. Assume 30%.
3. 3Purchases per buyer. Tea, a snack, parking, a shop purchase. Assume 1.8.
4. 4Share paid by UPI. Small-ticket payments in NCR markets are commonly UPI, but some pay cash or card. Assume 70%.

Demand-side merchant payments = 60,000 × 0.30 × 1.8 × 0.70 ≈ 22,700.

Add ticketing: assume 5% of the 60,000 buy a token or top up with UPI, about 3,000. Total ≈ 25,700.

Stage 4: the supply-side cross-check

Now count from the merchants.

1. 1Merchants within five minutes' walk. Street stalls, cafés, shops, parking attendants. Assume 400 that take payments.
2. 2Customers per merchant per weekday. Very uneven: a tea stall may serve hundreds, a jewellery shop a handful. Assume an average of 80.
3. 3UPI share. Same 70%.

Supply side = 400 × 80 × 0.70 = 22,400, plus 3,000 ticketing ≈ 25,400.

The two methods agree closely here. In practice they often differ by a factor of two, and that is fine — say which one you trust more and why. Here the supply side may be more reliable, because merchant counts are observable by walking the area, while "distinct people" is hard to pin down.

# Low / base / high ranges for the demand-side estimate (all assumptions)
scenarios = {
    "low":  dict(people=40_000, buy=0.20, purchases=1.5, upi=0.60, ticket=0.03),
    "base": dict(people=60_000, buy=0.30, purchases=1.8, upi=0.70, ticket=0.05),
    "high": dict(people=90_000, buy=0.40, purchases=2.2, upi=0.80, ticket=0.07),
}
for name, s in scenarios.items():
    merchant = s["people"] * s["buy"] * s["purchases"] * s["upi"]
    total = merchant + s["people"] * s["ticket"]
    print(f"{name:>4}: about {round(total, -3):,.0f} UPI payments per weekday")

The range runs from roughly 8,000 to 70,000. Presenting a base figure with a range shows you know which inputs are shaky.

Stage 5: sensitivity and the close

Which factor moves the answer most? In the demand method, "share who buy" and "distinct people" are both uncertain by a factor of two, and they multiply. The UPI share is actually the least uncertain input for small-ticket purchases.

Close with how you would measure it rather than guess it: count merchants on a walk of the area, stand at two stalls for an hour and count payments by mode, and scale by opening hours. Mention the weekend: shopping footfall probably rises while commuter footfall falls, so the total might be higher on Saturday — but that is itself an assumption to test.

A clean final line: "Base estimate roughly 25,000 to 26,000 UPI payments on a weekday, plausible range 8,000 to 70,000. The biggest uncertainty is how many passers-by buy anything; an hour of counting at a few stalls would narrow it."

Mistakes that cost marks

• Starting to multiply before defining what is being counted.
• Using a precise-sounding national statistic you half-remember. If you cannot source it, build from observable pieces instead.
• A single chain with no cross-check.
• Numbers with no reasons. "Assume 30%" needs "because most commuters walk straight through."
• Forgetting units — people versus transactions versus visits.

Six prompts to practise with the same plan

1. 1Daily auto-rickshaw rides starting from outside a large Noida metro station.
2. 2Cups of tea sold in a day across an IT park's food court.
3. 3Monthly food-delivery orders from one residential sector.
4. 4Number of two-wheeler service appointments per month at showrooms in one district.
5. 5Daily FASTag toll transactions at one expressway toll plaza.
6. 6Weekly visitors to a coaching institute's demo classes in a single market.

For each, write the scope sentence first, then one demand-side and one supply-side tree. A guesstimate practised this way is the same skill as scoping a real business question; the root-cause case for a sales dip uses the same habit of breaking a number into drivers. Estimation and business framing run through the Data Analyst programme, and if you are unsure whether analytics or data science suits you, read data analytics vs data science.