estimation-fermi_skill

Overview

This skill helps you produce fast, defensible order-of-magnitude estimates using Fermi-style decomposition. It’s designed for quick market sizing, feasibility checks, resource planning, and sanity checks when data is scarce or a directional answer suffices. The focus is on clear assumptions, bounding, and triangulation rather than false precision.

How this skill works

You clarify the target metric and decompose the unknown into smaller, estimable components (top-down, bottom-up, rate×time, density×area, or analogies). For each component you apply anchors, compute optimistic/pessimistic bounds, and run a sanity check. Finally, you triangulate with an alternate decomposition to validate the order-of-magnitude result.

When to use it

• Market sizing (TAM/SAM/SOM) for product or enter/expand decisions
• Quick infrastructure or staffing capacity planning and cost back-of-envelope
• Feasibility checks before investing in detailed analysis or experiments
• Sanity-checking strategic assumptions, pricing, runway, or unit economics
• Interview-style estimation questions or rapid decision triage

Best practices

• State all assumptions explicitly (units, timeframe, anchors) and round to 1–2 significant figures
• Decompose until each component is estimable from common knowledge or simple anchors
• Compute upper and lower bounds and report a range or confidence level, not a single precise number
• Triangulate with at least one alternate decomposition; investigate 10x disagreements
• Sanity-check results against known benchmarks, dimensional analysis, and extreme cases

Example use cases

• Estimate annual revenue opportunity for a B2B SaaS feature (users × ARPU × conversion)
• Roughly size servers needed for a new app (DAU × requests/user ÷ instance capacity)
• Determine headcount required to support X customers given tickets per customer and agent capacity
• Assess whether a new product could reach meaningful market scale before fundraising
• Quickly bound carbon impact of a proposed program (units affected × impact per unit × duration)