Estimation Questions
Estimation questions assess the ability to decompose unknown quantities into calculable components. The interviewer does not expect precise answers; they evaluate the reasoning process.
Purpose of Estimation Questions
PMs regularly make decisions with incomplete information:
- Feature prioritization without complete usage data
- Market sizing for new product opportunities
- Reasonableness checks on reported metrics
Estimation questions test comfort with uncertainty and structured reasoning.
Estimation Framework
| Step | Description |
|---|---|
| Decompose | Break the question into smaller, answerable components |
| Assume | State reasonable assumptions for each component |
| Calculate | Perform arithmetic with rounded numbers |
| Verify | Check the result against known reference points |
Approaches
Top-Down
Start with total population, narrow down through filters.
Example: Estimating email users
- US population: 330M
- Working age adults: 210M
- Percentage with jobs requiring email: 60%
- Percentage using Gmail: 70%
Bottom-Up
Start with individual unit, scale up.
Example: Estimating email volume
- Typical person sends 30 emails per day
- 150M workers in US
- 50% are heavy email users
Both approaches are valid. Using both and comparing results validates the estimation.
Reference Numbers
Population
| Region | Population |
|---|---|
| United States | 330 million |
| World | 8 billion |
| New York City | 8 million |
| California | 40 million |
US Demographics
| Category | Value |
|---|---|
| Households | 130 million |
| Adults | 260 million |
| Workers | 160 million |
| Registered vehicles | 280 million |
Time
| Metric | Value |
|---|---|
| Hours per year | 8,760 |
| Working hours per year | 2,000 |
| Work days per year | 250 |
Economics
| Metric | Value |
|---|---|
| Median US household income | $75,000 |
| Average coffee purchase | $5 |
| Average restaurant meal (2 people) | $50 |
| Median US rent | $1,500/month |
Worked Example: Uber Rides in NYC
Question: How many Uber rides occur in NYC per day?
Approach: Bottom-up from rider behavior
Calculation:
- NYC population: 8 million
- Demographics likely to use Uber regularly: 3 million (younger, urban, higher income)
- Percentage with app installed: 30% = 1 million potential riders
- Percentage active: 50% = 500K active riders
- Rides per active rider per week: 3
- Weekly rides: 1.5 million
- Daily rides: ~200K
Verification (supply-side):
- NYC for-hire vehicles: ~80,000
- Uber market share: ~50% = 40,000 Uber drivers
- Rides per driver per day: 5
- Daily rides: 200K
Both approaches converge, indicating reasonable estimation.
Worked Example: Pianos in Chicago
Question: How many pianos are in Chicago?
Approach: Sum of pianos by location type
Household pianos:
- Chicago population: 2.7M (use 3M)
- Persons per household: 2.5
- Households: 1.2M
- Percentage with piano: 4%
- Household pianos: 48,000
Institutional pianos:
- Schools: 800 total, 30% with pianos = 240
- Churches: 2,000 total, 20% with pianos = 400
- Other venues (hotels, bars, studios): ~500
Total: Approximately 50,000 pianos
Common Mistakes
| Mistake | Description |
|---|---|
| No visible reasoning | State assumptions and calculations explicitly |
| False precision | "47.3 million" suggests misunderstanding; use "roughly 50 million" |
| Unsupported assumptions | Every assumption requires brief justification |
| No verification | Compare final answer to known reference points |
| Silent thinking | Verbalize the reasoning process |
Evaluation Criteria
Interviewers assess:
| Criterion | Description |
|---|---|
| Structure | Logical decomposition of the problem |
| Assumptions | Reasonable estimates for each component |
| Communication | Clear explanation of reasoning |
| Adaptability | Ability to adjust when given new information |
| Self-correction | Identification and correction of errors |
Estimates within 2x of actual values are acceptable. Estimates off by 100x indicate reasoning errors that should be identified.
Practice Questions
Basic
- Annual haircuts in the US
- Daily text messages sent globally
- Daily departures from JFK airport
Intermediate
- Google's daily search advertising revenue
- Number of people currently flying
- YouTube's annual storage requirements
Advanced
- Tennis balls fitting in a room
- US pet insurance market size
- Traffic lights in Manhattan
Target completion time: under 5 minutes per question. Verify against available reference data.
Company-Specific Approaches
Google
Originated Fermi problem interviews. Focus shifted from abstract puzzles to business-relevant estimations. Research showed brain teasers poorly predicted job performance.
Amazon
Estimation questions relate to business operations: daily package deliveries, Prime Video storage costs. Expectation of familiarity with AWS pricing and business scale.
Uber/Lyft
Marketplace dynamics estimation. Driver supply needed for target ETAs, demand variation by time of day. Estimation connects directly to operational decisions.
Meta
Historical focus on user estimation questions. The underlying test: understanding of social product growth mechanics. Implausible viral growth estimates indicate weak product intuition.