Estimation Questions
This section provides frameworks and worked examples for estimation questions. For the underlying approach to estimation thinking, see the Estimation page.
PM Applications of Estimation
| Application | Description |
|---|---|
| Market sizing | TAM calculation for opportunity assessment |
| Capacity planning | Server/infrastructure requirements |
| Business cases | Revenue potential for feature funding decisions |
| Technical scoping | Data storage and processing requirements |
Estimation Framework
| Step | Action |
|---|---|
| 1. Clarify scope | Define exactly what is being counted |
| 2. Choose approach | Top-down or bottom-up |
| 3. Break into components | Identify smaller questions that compose the big question |
| 4. Estimate each piece | Make reasonable assumptions |
| 5. Sanity check | Compare to known reference points |
Estimation Approaches
Top-Down
Start with total population, narrow down by percentages.
| Step | Example |
|---|---|
| Total population | 330M (US) |
| Target segment % | 75% adults |
| Relevant behavior % | 20% who [behavior] |
| Product adoption % | 10% who would use |
Best for: Market sizing, TAM calculations, "how many people would use X"
Bottom-Up
Start with single unit, scale up.
| Step | Example |
|---|---|
| Single unit metric | 1 user does X per day |
| Per location | 500 users per store |
| Number of locations | 1,000 stores |
| Total | 500,000 X per day |
Best for: Operations questions, usage calculations, cost estimates
Recommendation: Run both approaches and check for convergence. Divergence indicates assumption error.
Worked Example: Cars in Seattle
Step 1: Clarify Scope
Questions to ask:
- Passenger vehicles only, or all vehicles?
- Seattle proper or metro area?
- Registered cars or cars present (including visitors)?
Assumptions: Passenger vehicles, Seattle metro area, registered cars.
Step 2: Choose Approach
Top-down (starting with population).
Step 3: Break Into Components
Seattle Metro Population x Percentage Adults x Car Ownership Rate = Number of Cars
Step 4: Estimate Each Component
| Component | Estimate | Reasoning |
|---|---|---|
| Seattle Metro Population | 4 million | Major US metro, tech hub |
| Adults (18+) | 75% | Typical US age distribution |
| Car ownership rate | 0.7 cars per adult | Urban area with transit, but still car-centric |
Calculation:
- 4,000,000 x 0.75 = 3,000,000 adults
- 3,000,000 x 0.7 = 2,100,000 cars
Estimate: Approximately 2 million cars
Step 5: Sanity Check
- US total: ~280 million cars for 330 million people (~0.85 per person)
- Seattle metro at US average: 4M x 0.85 = 3.4M
- Urban areas typically have lower car ownership
- 2M (0.5 per person) is reasonable for urban area
Worked Example: Gmail Cost Per Year
Step 1: Clarify Scope
- Total operating costs or marginal cost per user?
- Include development costs or infrastructure only?
Assumptions: Infrastructure operating costs only (development costs are sunk).
Step 2: Choose Approach
Bottom-up (cost per user x number of users).
Step 3: Break Into Components
| Cost Component | Factors |
|---|---|
| Storage | GB per user x cost per GB |
| Compute | Requests per user/day x cost per request |
| Bandwidth | Data transferred per user x cost per GB |
Step 4: Estimate Each Component
Users:
| Metric | Estimate | Reasoning |
|---|---|---|
| Gmail users | 1.8 billion | Google reports 1.5B+, assume growth |
| Active users | 1 billion | Not all check daily |
Storage:
| Metric | Estimate | Reasoning |
|---|---|---|
| Average mailbox size | 5 GB | 15 GB free, most use less |
| Storage cost | $0.02/GB/month | Cloud storage at scale |
| Annual storage cost | 1B x 5 x $0.02 x 12 = $1.2B |
Compute:
| Metric | Estimate | Reasoning |
|---|---|---|
| Email checks/day | 10 per active user | Mobile push + web |
| Cost per request | $0.0001 | Highly optimized |
| Daily compute | 1B x 10 x $0.0001 = $1M | |
| Annual compute | $365M |
Bandwidth:
| Metric | Estimate | Reasoning |
|---|---|---|
| Data per user/day | 5 MB | Emails + attachments |
| Cost per GB | $0.01 | Google's scale |
| Daily bandwidth | 1B x 5MB x $0.01/1000 = $50K | |
| Annual bandwidth | ~$20M |
Total: Approximately $1.6 billion per year
Step 5: Sanity Check
- $1.6B / 1.8B users =
$0.89 per user per year ($0.07/month) - Free product generating ad revenue: plausible
- Google total costs ~$100B; Gmail at ~1.5% is reasonable
Reference Numbers
Population & Demographics
| Metric | Value |
|---|---|
| World population | 8 billion |
| US population | 330 million |
| China/India population | 1.4 billion each |
| US households | 130 million |
| Smartphones worldwide | 6.5 billion |
| Internet users | 5 billion |
Economics
| Metric | Value |
|---|---|
| US GDP | $25 trillion |
| US median household income | $75,000 |
| Minimum wage (US) | ~$7-15/hour |
| Average rent (US) | $1,500/month |
Tech Metrics
| Metric | Value |
|---|---|
| Facebook DAU | 2 billion |
| Google searches per day | 8.5 billion |
| Amazon orders per day | 1.6 million |
| Average app session | 5-10 minutes |
| Cost per GB storage | $0.02/month |
| Cost per API call | $0.0001-0.001 |
Time & Units
| Metric | Value |
|---|---|
| Seconds in a year | ~32 million (approximately pi x 10^7) |
| Hours in a year | 8,760 |
| Working hours/year | 2,000 |
Estimation Techniques
Anchoring
Start with known reference point:
- "US population is 330M, so if 70% are adults..."
- "There are about 150K gas stations in the US, so for a city of 1M..."
Segmentation
Break population into meaningful groups with different behaviors:
- iPhone users (45%) vs. Android users (55%)
- Willingness to pay differs by segment
Bounding
Estimate high and low bounds, then narrow:
- "Cannot be less than 100K because..."
- "Cannot be more than 10M because..."
- "Somewhere in 500K-2M range"
Round Numbers
Use round numbers for mental math:
- 330M becomes 300M
- $47.50 becomes $50
- 17% becomes 20% (or 15%)
Common Errors
| Error | Example | Fix |
|---|---|---|
| No structure | "I guess... maybe 50,000?" | Write down framework |
| Wrong units | Mixing daily and annual | Label everything clearly |
| Dropping zeros | 1M x 1K = 1M (wrong: 1B) | Write out: 1,000,000 x 1,000 |
| Unrealistic assumptions | "Everyone uses X" | Use percentages, not absolutes |
| No sanity check | Ending with $10T market | Compare to known references |
| Going silent | Interviewer cannot help | Think out loud |
Question Categories
Market Sizing
| Question | Key Numbers |
|---|---|
| US online grocery market | US grocery ~$800B, online penetration ~10% |
| TAM for ride-sharing in Europe | EU population 450M, urban %, income levels |
| EV market in 5 years | Current 5%, growth rate 30%/year |
Operations
| Question | Key Numbers |
|---|---|
| Uber rides per day in NYC | NYC population 8M, 20% use rideshare |
| Storage needed for Netflix | 15K titles, average 3GB, multiple encodings |
| Google searches handled | 8.5B/day, 100K/second |
Physical Objects
| Question | Key Numbers |
|---|---|
| Golf balls in school bus | Bus ~40ft, ball ~1.5 inch |
| Piano tuners in Chicago | Pianos per household, tuning frequency |
| Gas stations in US | 150K (useful anchor) |
Interview Best Practices
| Practice | Description |
|---|---|
| Think out loud | Process matters more than answer |
| State assumptions | "I'm assuming average user checks email 10 times per day" |
| Show math | Write step by step |
| Accept uncertainty | "Between 10-20%" is acceptable |
| Round aggressively | 17.3% to 20% does not materially change answer |
| Sanity check | "Let me verify by comparing to..." |
Practice Questions
Easy (5 minutes each)
- Pizzas eaten in US per year
- iPhone cases sold annually
- Market size for dog food in US
Medium (10 minutes each)
- Revenue Spotify generates per user per year
- WhatsApp messages sent globally per day
- Market size for online education in US
Hard (15 minutes each)
- Cost to build a Spotify competitor
- YouTube's annual bandwidth costs
- TAM for autonomous delivery vehicles in next 10 years