## Core Insight In estimation problems, the method matters more than the answer. Interviewers evaluate how you decompose a vague question into tractable sub-problems, not whether your final number is correct. A wrong answer with a clear, reproducible process is better than a correct answer arrived at by guessing. ## The Method 1. **Round aggressively** — "99,987 / 9.1" becomes "100,000 / 10." Precision is not expected and wastes time. 2. **Write down assumptions** — make them explicit so they can be challenged, referenced, and adjusted. An assumption written down is a parameter; an assumption in your head is a guess. 3. **Label units** — "5" is ambiguous. "5 MB" is clear. When you write "5," does it mean 5 KB or 5 MB? Labeling prevents order-of-magnitude errors. 4. **Know the reference tables** — powers of 2 (KB/MB/GB/TB), latency numbers (memory ~100ns, disk seek ~10ms, network round-trip ~150ms cross-continent), availability nines (99.99% = 52 min/year downtime). 5. **Estimate common quantities** — QPS, peak QPS (~2x average), storage per day/year, number of servers needed. ## Why Process Wins - **Reproducible**: If assumptions change, you can re-derive. A memorized answer is useless when parameters shift. - **Debuggable**: When the final number seems wrong, you can trace back to which assumption is off. - **Communicative**: The interviewer sees your reasoning, not just a number. This is the actual skill being tested. ## Cross-Domain Applications - **Business planning**: Revenue projections with explicit assumptions beat precise spreadsheets with hidden ones. Investors fund the reasoning, not the number. - **Project estimation**: Breaking "how long will this take?" into sub-tasks with individually estimable durations, then summing. Process reveals which parts are uncertain. - **Capacity planning**: SRE capacity forecasting uses the same decomposition — current traffic × growth rate × headroom multiplier. The components matter more than the product. ## Source - [[Back-of-the Envelope Estimation]] (Alex Xu, System Design Interview Ch 2)