the brief → sign in
15 min left 0%

estimation & guesstimate cases

The general approach for guesstimate questions is that you start with a general formula. The interviewer is assessing your structural thinking, not the number.
From a Pragmatic Leaders interview prep session with Fernando Delgado

Estimation questions are not math tests. They are thinking tests dressed up as math.

I have reviewed thousands of guesstimate answers from PMs across India. The pattern is consistent: candidates who build a structure first and fill numbers second always outperform candidates who start multiplying immediately. The number is almost irrelevant. The decomposition is everything.

This page walks through 13 fully worked estimation cases — market sizing, volume estimates, revenue projections, unit economics — all solved step by step with India-specific numbers. Study the scaffolding, not the arithmetic.

Before you start: the toolkit

Every estimation follows the same skeleton. Internalize it before reading the cases.

StepWhat you doWhy it matters
1. ClarifyScope the question. Geography? Time period? Definition?”Market size of EdTech” could mean Rs 5,000 crore or Rs 50,000 crore depending on what you include
2. Pick your approachTop-down (universe → filter) or bottom-up (unit → scale)Declare it. Out loud. Before you calculate.
3. DecomposeBreak into 4-6 independent variables you can estimate separatelyThis is where the real skill lives
4. Estimate each variableState assumption, give range, pick a number, move onDo not get stuck debating one number for 3 minutes
5. CalculateMultiply through. Talk while you compute.The interviewer needs to follow your math in real time
6. Sanity checkCross-check with a second method or a known reference pointConvergence is the strongest signal you can give

Your India-specific anchors — memorize these:

  • India population: ~1.4 billion. Urban: ~35% (~50 crore). Top 8 metros: ~10 crore total.
  • Bangalore: ~1.3 crore. Mumbai metro: ~2.5 crore. Delhi NCR: ~3 crore.
  • Average household size: 3.5 (urban), 4.5 (rural).
  • Smartphone penetration: ~75% urban, ~45% rural. ~80 crore internet users.
  • Average urban household income: Rs 40,000-50,000/month.
  • UPI transactions: ~14 billion/month (~47 crore/day).

Case 1: TAM for online grocery in India

// scene:

Final round at a quick-commerce startup. The VP Product leans forward.

VP Product: “What is the total addressable market for online grocery in India? Give me a number and show me how you got there.”

Candidate: “Before I start — are we talking about all grocery including kirana delivery, or specifically app-based platforms like Blinkit, Zepto, BigBasket?”

VP Product: “App-based. Organized online grocery.”

Candidate: “And TAM as in the theoretical maximum if everyone who could order online did, or the realistic serviceable market in the next 2-3 years?”

VP Product: “Realistic serviceable. SAM, if you prefer.”

Two questions. The problem just went from 'all grocery in India' to 'app-based organized online grocery in serviceable geographies over the next 2-3 years.' That is a 10x narrower — and 10x more answerable — question.

// tension:

Scope before you solve. Always.

Top-down approach

  1. Total grocery market in India: Rs 35-40 lakh crore annually ($420-480B). Use Rs 38 lakh crore.
  2. Urban share: ~60% of grocery spend is urban = Rs 22.8 lakh crore.
  3. Metros + Tier-1 cities (serviceable geography): ~40% of urban spend = Rs 9.1 lakh crore.
  4. Addressable households: Those with smartphones, internet, and willingness to order grocery online. ~35% of metro/Tier-1 households today = Rs 3.2 lakh crore addressable spend.
  5. Online penetration of addressable: Currently ~8-10% of grocery spend from these households goes online. In 2-3 years, realistically 15-18%. Use 16%.
  6. SAM: Rs 3.2 lakh crore × 0.16 = Rs 51,000 crore ($6.1B).

Bottom-up cross-check

  1. Online grocery-active households in metros + Tier-1: ~2.5 crore households.
  2. Monthly spend per household on online grocery: ~Rs 3,500 (mix of quick-commerce Rs 500-700/week households and monthly BigBasket Rs 2,000-4,000 basket households). Average across active and light users: Rs 1,700/month.
  3. Annual: 2.5 crore × Rs 1,700 × 12 = ~Rs 51,000 crore.

Convergence at Rs 51,000 crore. That is your SAM.

What makes this answer work: The candidate did not say “India’s grocery market is Rs 38 lakh crore.” That is TAM, and it is meaningless for a quick-commerce startup. They filtered to the geography and customer segment the company can actually serve.

Case 2: Daily auto-rickshaw rides in Hyderabad

Top-down

  1. Hyderabad metro population: ~1 crore.
  2. Daily commuters (working age, leaving home): ~55% = 55 lakh.
  3. Mode share for auto-rickshaws: ~12% (rest: two-wheelers 35%, bus/metro 25%, car 15%, walk/cycle 8%, cab 5%).
  4. Daily commute trips by auto: 55 lakh × 0.12 = 6.6 lakh one-way trips.
  5. Round trips: 6.6 lakh × 2 = 13.2 lakh commute-related auto trips.
  6. Non-commute trips (hospital, market, station): Add ~40% = 13.2 lakh × 1.4 = ~18.5 lakh auto rides/day.

Bottom-up

  1. Registered auto-rickshaws in Hyderabad: ~1.5 lakh (RTA data).
  2. Active on any given day: ~70% = ~1.05 lakh.
  3. Average rides per auto per day: ~15-18 (morning rush, afternoon lull, evening rush). Use 16.
  4. Total: 1.05 lakh × 16 = ~16.8 lakh rides/day.

Range: 16.8-18.5 lakh. The gap is small. Top-down slightly overestimates because not all non-commute trips use autos (some use bikes). Bottom-up might undercount because shared autos carry multiple passengers per trip. Both land in the same zone.

Case 3: Revenue of BookMyShow per month

Decomposition

  1. Monthly active transacting users in India: ~1.5 crore (BookMyShow claims 100M+ app downloads; active monthly transactors are much smaller — roughly 15% of MAU).
  2. Average transactions per user per month: ~1.2 (most people book once; some book 2-3 times for movies + events).
  3. Average ticket value: Rs 350 (mix of Rs 200-250 single screens and Rs 500-700 multiplexes, weighted toward multiplexes).
  4. Gross merchandise value (GMV): 1.5 crore × 1.2 × Rs 350 = Rs 630 crore/month.
  5. BookMyShow take rate: ~8-12% on movie tickets (they keep the convenience fee). Use 10%.
  6. Net revenue from movies: Rs 630 crore × 0.10 = Rs 63 crore/month.
  7. Events + sports + streams (higher margin, ~20-25% take rate): Add ~30% to net revenue = Rs 63 crore × 1.3 = ~Rs 82 crore/month.

Sanity check: BookMyShow parent company Bigtree reported ~Rs 900-1,000 crore annual net revenue in recent filings. Rs 82 crore × 12 = Rs 984 crore. Right in range.

Case 4: Number of CCTV cameras in Mumbai

This is a physical counting problem. No revenue, no users — just objects in space.

Bottom-up

  1. Mumbai area: ~600 sq km (municipal limits).
  2. Major road length: ~2,000 km of arterial and collector roads.
  3. CCTV density on major roads: ~1 camera every 200 meters (Mumbai police has been aggressive) = 10,000 cameras on roads.
  4. Residential societies: ~50,000 housing societies in Mumbai. Average 4-6 cameras each = ~2.5 lakh cameras.
  5. Commercial establishments (malls, offices, shops): ~3 lakh commercial units with cameras, average 3 cameras = ~9 lakh cameras.
  6. Government/institutional (railway stations, airports, hospitals): ~50,000 cameras.
  7. Total: 10,000 + 2.5 lakh + 9 lakh + 50,000 = ~12.1 lakh CCTV cameras.

There is no easy top-down for physical counting problems. Instead, sanity-check against a ratio: ~12 lakh cameras for ~2 crore people = ~6 cameras per 100 people. Global benchmarks for surveilled cities range from 2-15 per 100 people. Mumbai at 6 is plausible — high for India, moderate globally.

Case 5: Annual revenue of Zerodha

// thread: #pm-prep — PM interview prep group
Ananya Got asked to estimate Zerodha's annual revenue in a fintech interview. Blanked on where to start.
Ravi Start from their user base and work through what they charge. It is a volume game.
Ananya But they do not charge brokerage on equity delivery. So where does the money come from?
Ravi F&O. That is the entire business. Rs 20 per order, and India trades a LOT of options.

Decomposition

  1. Zerodha active clients: ~75 lakh (publicly reported).
  2. What percentage trades F&O? ~15-20% actively trade derivatives in any month. Use 17% = ~12.75 lakh active F&O traders.
  3. Average F&O orders per active trader per month: ~40-50 (options traders are frequent). Use 45.
  4. Monthly F&O orders: 12.75 lakh × 45 = ~5.7 crore orders.
  5. Revenue per F&O order: Rs 20 flat brokerage.
  6. Monthly F&O brokerage: 5.7 crore × Rs 20 = Rs 114 crore.
  7. Other revenue (subscription fees for Kite/Console, mutual fund commissions, interest on idle funds): Add ~35% = Rs 114 crore × 1.35 = Rs 154 crore/month.
  8. Annual: Rs 154 crore × 12 = ~Rs 1,850 crore.

Sanity check: Zerodha publicly reported ~Rs 2,000-2,200 crore revenue in recent years. The estimate is in the right ballpark. The gap likely comes from underestimating the “other income” bucket (interest income alone can be substantial for a brokerage holding client funds).

Case 6: How many weddings happen in India per year?

Top-down

  1. India population: 1.4 billion.
  2. Marriageable age population (20-35): ~30% = 42 crore.
  3. Annual marriage rate: ~8-9 marriages per 1,000 population (census-derived). Use 8.5 per 1,000.
  4. Total marriages/year: 1.4 billion × 8.5/1,000 = ~1.2 crore weddings per year.

Bottom-up

  1. Average wedding involves 2 people getting married (obvious, but important — do not double-count).
  2. People who get married each year: India’s median first-marriage age is ~23 for women and ~26 for men. The 20-30 age cohort is ~28 crore. If average marriage happens once across a ~10-year window, roughly 1/10th marry each year = ~2.8 crore individuals = ~1.4 crore weddings.

Range: 1.2-1.4 crore. Government estimates put it at ~1 crore to 1.2 crore. Close enough. The bottom-up slightly overcounts because not everyone marries in that decade (some marry later, some do not marry at all).

This matters for PMs because the Indian wedding market is ~Rs 10 lakh crore ($120B). If you are sizing a wedding-tech product, this population number is your base.

Case 7: Monthly revenue of a single Starbucks store in Koramangala

A revenue estimation at the unit level. Useful for any interview where the question is about a single location, restaurant, or outlet.

  1. Operating hours: 8 AM to 11 PM = 15 hours.
  2. Seating capacity: ~60-70 seats in a typical Koramangala Starbucks. Use 65.
  3. Table turnover: Morning rush (8-10 AM): 2 turns. Midday (10 AM-4 PM): 1.5 turns. Evening (4-8 PM): 2.5 turns. Late (8-11 PM): 1 turn. Weighted average across the day: ~1.7 turns.
  4. Customers per day from seating: 65 × 1.7 = ~110.
  5. Takeaway and delivery: Add ~60-70% more footfall that does not sit = ~110 × 1.65 = ~180 additional.
  6. Total daily customers: 110 + 180 = ~290.
  7. Average ticket: Rs 380 (Koramangala crowd skews premium — lattes, not filter coffee).
  8. Daily revenue: 290 × Rs 380 = ~Rs 1.1 lakh.
  9. Monthly (30 days): Rs 1.1 lakh × 30 = ~Rs 33 lakh/month.

Sanity check: Tata Starbucks averages ~Rs 3.2 crore/store/year across all India stores, which is Rs 26-27 lakh/month. Koramangala is premium, so Rs 33 lakh is reasonable — a top-quartile store.

Case 8: TAM for pet care in India

Top-down

  1. Indian households: ~30 crore.
  2. Pet-owning households: ~3-4% (India is much lower than the US at ~67%). Use 3.5% = ~1.05 crore households.
  3. Average annual spend per pet-owning household: Rs 15,000-20,000 (food Rs 8,000-10,000, vet Rs 3,000-5,000, grooming/accessories Rs 2,000-5,000). Use Rs 18,000.
  4. Total market: 1.05 crore × Rs 18,000 = Rs 18,900 crore ($2.3B).

Online-addressable SAM

  1. Pet owners in metros + Tier-1 (smartphone, willing to order online): ~60% of pet-owning households = ~63 lakh.
  2. Online share of spend: ~20% currently (growing fast) = Rs 18,000 × 0.20 = Rs 3,600/household.
  3. Online pet care SAM: 63 lakh × Rs 3,600 = ~Rs 2,270 crore.

Industry reports peg India’s pet care market at Rs 15,000-22,000 crore depending on definitions. The estimate sits in range.

Case 9: Number of potholes in Bangalore

An absurd-sounding question. But the decomposition is the same as any other estimation.

  1. Total road length in Bangalore: ~14,000 km (BBMP data).
  2. Percentage of roads in poor condition: ~35-40% (anyone who has driven in Bangalore knows this is conservative). Use 37%.
  3. Roads in poor condition: 14,000 × 0.37 = ~5,200 km.
  4. Average potholes per km of poorly maintained road: ~8-12. Use 10.
  5. Potholes on bad roads: 5,200 × 10 = 52,000.
  6. Even “good” roads have some potholes: Remaining 8,800 km × 1.5 per km = 13,200.
  7. Total: 52,000 + 13,200 = ~65,000 potholes.

Sanity check: BBMP reportedly filled ~15,000-20,000 potholes in a recent monsoon season alone, and Bangaloreans will tell you they barely made a dent. 65,000 at any given time passes the smell test.

Case 10: Unit economics of a Swiggy delivery

This is a cost decomposition, not a revenue estimate. Interviewers use it to test whether you understand marketplace economics.

ComponentEstimateReasoning
Delivery partner payoutRs 35-45 per orderBase Rs 25 + per-km Rs 5-8 for avg 3 km delivery. Use Rs 40.
Payment gateway feeRs 5-8~1.5-2% of avg order value Rs 350. Use Rs 6.
Tech infrastructure per orderRs 2-3Cloud + ML/matching amortized across millions of orders. Use Rs 2.5.
Customer support allocationRs 3-5~2% of orders need support, each costing Rs 150-200. Amortized: Rs 3.5.
Customer acquisition cost (amortized)Rs 15-25CAC Rs 300-500, LTV ~20 orders. Rs 400/20 = Rs 20.
Packaging subsidyRs 2-3Partial subsidy for branded bags. Use Rs 2.5.
Total cost per order~Rs 74.5
Revenue per order~Rs 65-75Commission ~22% of Rs 350 = Rs 77, minus discounts/promos ~Rs 8-10 = Rs 67-69.

Contribution margin: Roughly break-even to slightly negative (Rs -5 to +5 per order). This is why Swiggy pushes Instamart and Swiggy One — bundling multiple order types across a single delivery fleet improves unit economics.

Case 11: How many ATMs are there in India?

Top-down

  1. RBI reports: India had ~2.15 lakh ATMs at peak. But this is a “how would you estimate it” question, not a “recall the number” question. So build it from scratch.
  2. India bank branches: ~1.5 lakh (across all scheduled commercial banks).
  3. ATMs per bank branch: Historically ~1.5-2 (some branches have 2 ATMs, many standalone ATMs exist). Use 1.5.
  4. ATMs at bank branches: 1.5 lakh × 1.5 = ~2.25 lakh.
  5. Standalone/off-site ATMs: RBI data suggests roughly 40% of ATMs are off-site. If on-site is 60% and equals 2.25 lakh, total = 2.25 lakh / 0.6 = ~3.75 lakh.

That overshoots. The problem: my ATMs-per-branch ratio included some standalone ATMs already. Adjust to 1.2 ATMs per branch = 1.8 lakh on-site. Total = 1.8/0.6 = 3 lakh. Still high.

Bottom-up cross-check:

  1. Urban India: ~50 crore people. RBI target: 1 ATM per 1,000-1,500 urban population. Use 1 per 1,200 = ~4.2 lakh. But India has not met that target — actual is maybe 60% = 2.5 lakh urban ATMs.
  2. Rural India: ~90 crore people. ATM density is very low — 1 per 10,000-15,000 people = ~7,000-9,000. Use 8,000.
  3. Total: ~2.5 lakh + 8,000 = ~2.6 lakh ATMs.

Known data: RBI reports ~2.1-2.2 lakh ATMs as of 2025 (declining due to UPI). The estimate slightly overshoots, which makes sense — I assumed higher density than the post-UPI reality.

The learning here: when your estimate overshoots a known number, the interesting question is why. In this case, UPI adoption has been actively shrinking the ATM network. A good candidate would name that structural shift.

Case 12: Revenue of Nykaa per quarter

// scene:

Case study round at an e-commerce company. The interviewer has been nodding along as the candidate builds out the estimation.

Candidate: “Nykaa has about 3-3.5 crore annual transacting customers. In any given quarter, maybe 40-45% transact — so roughly 1.4 crore quarterly active buyers.”

Interviewer: “Why 40-45%? Walk me through that.”

Candidate: “Beauty products have a repurchase cycle of 2-3 months for consumables like skincare, longer for makeup. So in a quarter, roughly half the annual base makes at least one purchase. I am using 40% to be conservative.”

The interviewer nods. The candidate justified the assumption — that is what separates a good estimate from a guess.

// tension:

Every number needs a 'because.' If you cannot justify an assumption, flag it as a risk and move on.

Full decomposition

  1. Quarterly active buyers: ~1.4 crore.
  2. Average order frequency per quarter: ~1.5 orders (some buy once, some buy monthly during sale events).
  3. Average order value (AOV): Rs 1,200-1,500 for beauty/personal care. Fashion is higher (~Rs 1,800) but smaller share. Blended: Rs 1,350.
  4. Quarterly GMV: 1.4 crore × 1.5 × Rs 1,350 = ~Rs 2,835 crore.
  5. Net revenue: Nykaa operates owned-inventory (BPC) + marketplace (fashion). Blended take rate ~40-45% (they buy and resell most BPC products, so revenue is closer to GMV than a pure marketplace). Use 42%.
  6. Quarterly net revenue: Rs 2,835 crore × 0.42 = ~Rs 1,190 crore.

Sanity check: Nykaa reported quarterly revenue of Rs 1,200-1,600 crore in recent quarters. The base estimate of Rs 1,190 crore is at the lower end because I excluded offline retail (Nykaa has 150+ physical stores contributing ~10% of revenue). Add that: Rs 1,190 × 1.1 = Rs 1,310 crore. Comfortably in range.

Case 13: How many Instagram Reels are uploaded in India per day?

Top-down

  1. Instagram MAU in India: ~35 crore (India is Instagram’s largest market).
  2. DAU: ~55% of MAU = ~19 crore.
  3. Percentage of DAU who create content (vs. consume): ~3-5% (the 1/9/90 rule — 1% create, 9% engage, 90% lurk; Reels has pushed creation higher). Use 4%.
  4. Daily creators: 19 crore × 0.04 = ~76 lakh.
  5. Reels vs. other formats: Of creators, ~60-70% now post Reels (vs. stories, static posts). Use 65%.
  6. Daily Reel creators: 76 lakh × 0.65 = ~49 lakh.
  7. Average Reels per creator per day: ~1.2 (most post one; some influencers post 2-3).
  8. Daily Reels uploaded from India: 49 lakh × 1.2 = ~59 lakh Reels/day.

There is no clean bottom-up for content creation at this scale. Sanity check instead: Instagram reportedly crosses 200 crore Reels plays/day globally. India at ~35% of user base implies ~70 crore plays/day in India. If each Reel gets an average ~12-15 plays (most Reels get very few), 70 crore / 12 = ~58 lakh Reels. Consistent.

The decision point

// interactive:
Your Estimation is Challenged

You have just finished estimating the TAM for online pharmacy in India. You arrived at Rs 14,000 crore. The interviewer looks at you and says: 'That feels high. Convince me, or revise it.'

How do you respond?

Common patterns across all 13 cases

After working through these, notice what repeats:

  1. Scope first. Every good answer starts with clarifying the question. “All India or metros?” “All grocery or app-based?” “Revenue or GMV?” This is not stalling — it is the single highest-impact move.

  2. Declare your method. Say “I will use a top-down approach starting from X” before you calculate. It gives the interviewer a map to follow.

  3. State every assumption. “I am assuming 70% smartphone penetration in urban India.” If you do not say it, the interviewer does not know whether you picked that number thoughtfully or randomly.

  4. Converge from two directions. Top-down and bottom-up arriving at the same number is the strongest possible close. When they diverge, the conversation about why they diverge is often more interesting than the number itself.

  5. Use India-specific anchors. Do not use US numbers scaled down. India’s urban household size, income distribution, smartphone penetration, and platform adoption curves are different. Know your base numbers for India.

  6. Round aggressively. Rs 51,247 crore is worse than “roughly Rs 50,000 crore.” False precision makes you look like you do not understand what estimation means.

// exercise: · 20 min
Build your estimation muscle

Pick three of these and solve each using both top-down and bottom-up. Write them on paper. Time yourself — 5 minutes per question, which is what you will get in an interview.

  1. How many two-wheelers are sold in India per month?
  2. What is the annual revenue of Haldiram’s?
  3. How many daily active users does CRED have?
  4. How many litres of milk does Amul sell per day?
  5. What is the TAM for SaaS tools in India?
  6. How many flights take off from Mumbai airport per day?

After each one:

  • Did your top-down and bottom-up converge within 2x? If not, which assumption was off?
  • Could you explain every number to an interviewer who keeps asking “why?”
  • Did you finish within 5 minutes?

The goal is not accuracy. The goal is building a decomposition reflex that activates before you start multiplying.

// learn the judgment

A BigBasket PM is asked to estimate the number of SKUs that can be profitably delivered in under 10 minutes in Bengaluru. The data team says they need 3 weeks to build the model. The category head wants the answer in 2 days for a board presentation.

The call: Do you tell the category head you need 3 weeks, or do you build a rough estimate in 2 days?

// practice for score

A BigBasket PM is asked to estimate the number of SKUs that can be profitably delivered in under 10 minutes in Bengaluru. The data team says they need 3 weeks to build the model. The category head wants the answer in 2 days for a board presentation.

The call: Do you tell the category head you need 3 weeks, or do you build a rough estimate in 2 days?

0 chars (min 80)

Where to go next

estimation & guesstimate cases 0%
15 min left