estimation & guesstimate questions

Every PM interview in India includes some version of this question: “How many X happen in Y?” The interviewer does not care about the number. They care about how you decompose the problem, what assumptions you make, and whether those assumptions are reasonable.

I have watched thousands of PMs attempt these questions. The ones who fail share a common pattern: they start guessing numbers immediately. The ones who pass share a different pattern: they build a structure first, then fill in the numbers.

This page gives you that structure.

The scene

The candidate’s number might even be close. But the interview is already lost. The interviewer was not asking for a number — they were asking for a thinking process.

Two approaches, same question

There are exactly two ways to solve any estimation question: top-down and bottom-up. Every guesstimate is one of these, and the best answers use both as a cross-check.

Top-down: start with the universe, filter down

You begin with the largest relevant number and apply successive filters to narrow it.

“How many Zomato deliveries happen in Bangalore per day?” — Top-down approach:

1. Bangalore population: ~1.3 crore (13 million)
2. Household size: Average 3.5 people per household = ~37 lakh households
3. Internet-connected, smartphone-owning households: ~70% in urban Bangalore = ~26 lakh
4. Households that have ever ordered food online: ~50% = ~13 lakh
5. Active monthly orderers (ordered in last 30 days): ~40% of those = ~5.2 lakh
6. Average orders per active household per month: ~6 (roughly every 5 days)
7. Monthly orders: 5.2 lakh × 6 = ~31 lakh
8. Daily orders: 31 lakh ÷ 30 = ~1 lakh orders per day

Now, Zomato has roughly 50-55% market share vs Swiggy in most metros. So Zomato-specific: ~50,000-55,000 deliveries per day.

Bottom-up: start with one unit, scale up

You begin with the smallest observable unit and multiply outward.

Same question — Bottom-up approach:

1. How many restaurants does Zomato have in Bangalore? Roughly 15,000-20,000 listed; maybe 8,000 active on delivery.
2. Average orders per active restaurant per day: Varies hugely. A popular Koramangala biryani place might do 80-100 orders. A small bakery might do 5. Average across all restaurants: ~7 orders/day.
3. Total: 8,000 × 7 = ~56,000 deliveries per day

Both approaches land in the same range: 50,000-60,000. When top-down and bottom-up converge, you have a credible estimate. When they diverge wildly, one of your assumptions is wrong — and finding which one is where the interesting conversation happens.

What the interviewer is actually grading

The number itself matters least. I have seen candidates get hired with estimates that were 3x off because their structure was clean and their assumptions were sensible. I have seen candidates get rejected with accurate numbers because they could not explain how they got there.

Common estimation categories

Every guesstimate you will face in an Indian PM interview falls into one of these categories. Knowing the category tells you which approach and which anchor numbers to start with.

The starting anchor matters. For any India-specific question, you need a small set of numbers in your head:

• India population: ~1.4 billion. Urban: ~35%. Top 8 metros: ~10 crore total.
• Bangalore: ~1.3 crore. Mumbai: ~2 crore (metro region: ~2.5 crore). Delhi NCR: ~3 crore.
• Average household size: 3.5 (urban), 4.5 (rural).
• Smartphone penetration: ~75% urban, ~45% rural.
• Internet users in India: ~80 crore. Daily active: ~55 crore.
• Average monthly household income (urban): ~Rs 40,000-50,000.

Memorize these. They are the starting point for every guesstimate.

Worked example 1: PhonePe transactions per day in India

Top-down approach:

1. India’s UPI processed ~14 billion transactions in a recent month. That is ~47 crore transactions/day.
2. PhonePe has roughly 47-48% UPI market share.
3. PhonePe daily transactions: 47 crore × 0.47 = ~22 crore transactions per day.

Bottom-up approach:

1. PhonePe has ~50 crore registered users. Monthly active users: ~35 crore (70% MAU/registered ratio).
2. Daily active users: ~15 crore (roughly 43% of MAU — reasonable for a payments app).
3. Average transactions per DAU per day: ~1.5 (some people transact multiple times; many open the app but only transact once or not at all).
4. Daily transactions: 15 crore × 1.5 = ~22.5 crore.

Both converge around 22 crore. That convergence is your strongest signal in an interview.

Worked example 2: How many Ola/Uber rides happen in Delhi NCR daily?

Top-down:

1. Delhi NCR population: ~3 crore.
2. Working-age population (18-60): ~60% = 1.8 crore.
3. People who commute daily: ~70% = 1.26 crore.
4. Mode split — cab/ride-hailing share of commutes: ~8% (most use metro, bus, two-wheeler, auto).
5. Daily ride-hailing commute trips: 1.26 crore × 0.08 = ~10 lakh trips.
6. But commute is only ~40% of ride-hailing usage. Add non-commute (airport, hospital, shopping, nights out): 10 lakh ÷ 0.4 = ~25 lakh total rides/day.

Bottom-up:

1. Estimated active ride-hailing drivers in Delhi NCR: ~1.5 lakh (Ola + Uber combined).
2. Average rides per driver per day: ~12-15 (accounting for downtime, surge hours, idle periods). Use 14.
3. Total rides: 1.5 lakh × 14 = ~21 lakh rides/day.

Range: 21-25 lakh. Close enough. In the interview, state both numbers and explain why the gap exists (top-down might overcount non-commute; bottom-up might undercount drivers).

Worked example 3: Annual revenue of Starbucks in India

Top-down:

1. Starbucks India has ~420 stores (as of early 2026, through Tata Starbucks JV).
2. Average customers per store per day: A well-located store in Koramangala or Connaught Place does 300-400. A mall store might do 150-200. Average across all stores: ~250.
3. Average ticket size: Rs 350 (one coffee + maybe a snack).
4. Daily revenue per store: 250 × 350 = Rs 87,500.
5. Annual revenue per store: Rs 87,500 × 365 = ~Rs 3.2 crore.
6. Total annual revenue: 420 × 3.2 crore = Rs 1,340 crore ($160M).

Sanity check: Tata Consumer Products reports Starbucks India revenue around Rs 1,200-1,400 crore in their filings. The estimate is in range.

Notice I used only one approach here but added a sanity check against a known reference. In an interview, that cross-reference is almost as powerful as a second estimation method.

The five mistakes that kill your answer

1. Not stating assumptions. Every time you use a number, say where it came from. “I’m assuming Bangalore has 1.3 crore people” is fine. Just picking 1.3 crore without saying it is an assumption looks like you are guessing.

2. False precision. Saying “there are 1,27,43,892 smartphone users in Bangalore” is worse than saying “roughly 90 lakh.” Round numbers show you understand that this is an estimate, not a census.

3. Forgetting to segment. Treating all of India as one homogeneous market is the fastest way to get an absurd number. Urban and rural are different. Metro and Tier-2 are different. Working and non-working populations order food at different rates. Segment.

4. Not talking through the math. The interviewer cannot see inside your head. Talk while you calculate. “So 37 lakh households times 50% gives me about 18-19 lakh, let me round to 18 lakh.” This lets them follow your logic and also gives them a chance to nudge you if an assumption is off.

5. Stopping after one approach. If you have time, always cross-check with a second method. When both methods converge, say so — it is the strongest possible close. When they diverge, acknowledge it and explain which assumptions might be off. That analysis is more impressive than getting the right number.

The branching scenario

Where to go next

• Combine estimation with case analysis: Case Study Frameworks
• Understand what interviewers are really testing: What Interviewers Actually Evaluate
• See the full landscape of PM interview types: Interview Types