Estimation Language
Give professional, calibrated estimates in English — from hedged time commitments in standups to order-of-magnitude back-of-envelope calculations in system design discussions.
Intermediate 6 exercises
Time Estimates
Give hedged estimates, qualify uncertainty, and revise estimates with professional language.
Start → Advanced 6 exercises
Back-of-Envelope Estimates
Narrate order-of-magnitude calculations: DAU, storage, throughput, and scale reasoning.
Start → Intermediate 6 exercises
Reading SLAs & SLOs
Interpret uptime percentages, downtime budgets, and service level terminology.
Start → Intermediate 6 exercises
Performance Metrics Language
Narrate latency improvements, error rates, throughput gains, and percentile statistics.
Start → Advanced 6 exercises
Capacity Planning Language
Express resource requirements, scaling assumptions, and infrastructure projections clearly.
Start →Estimation phrase reference
Hedged time estimates
- Rough estimate: "I'd say around 2–3 days, but that could change once I look at the data layer."
- Qualifier: "assuming no surprises / assuming the API is stable"
- Range: "somewhere between 3 and 5 days depending on complexity"
- Revised estimate: "I need to revise my estimate — it's looking more like a week now that we've scoped the auth changes"
- Under pressure: "I can give you a number, but I'd want to caveat it — I haven't seen the spec yet"
Back-of-envelope language
- Assume: "let's say 10M DAU, so roughly 115 requests/second on average"
- Order of magnitude: "we're talking about tens of gigabytes, not hundreds"
- Rounding: "rounding up to be safe" / "conservative estimate"
- Per-unit: "if each user uploads ~5 photos at 2MB, that's 10MB per user per day"
- Extrapolate: "at 5× growth over 12 months, we'd need…"
SLA & metrics vocabulary
- Uptime clause: "99.9% uptime = ~8.7 hours downtime/year"
- Error budget: "we've consumed 40% of our monthly error budget"
- Latency: "median latency" (P50) / "tail latency" (P99) / "P95 latency"
- Throughput: "X requests per second" / "Y transactions per minute"
- Improvement: "reduced P99 latency from 1.2s to 280ms — a 4.3× improvement"