From prod MCP feedback (2026-06-10, feedback_events #7): every sizing should output THREE time-anchored figures, each with its own evidentiary grade and expiry, never collapsed into one CAGR.
- Realized TAM (most recent completed period) — booked, primary-sourced; bar HIGH (regulator/filing).
- Run-rate TAM (~1–2y) — sum of exit run-rates at snapshot; bar MEDIUM (dated press/vendor disclosures).
- Prize TAM (~5y+) — value pool × capture rate; FRAMEWORK-ONLY, explicitly labeled, never a forecast.
Guardrails (from the feedback's suggested_value JSON): report three figures separately with grades; no CAGR blending across grades; decisions must declare which figure they price against; units consistent per figure (GAAP vs ARR vs annual potential); separate expiries (run-rate expires ~quarterly in hypergrowth).
Worked example given: AI coding tools — realized 2025 ~$3-4B (B), run-rate 2026 ~$7-10B (B-), prize $50-150B (framework-only).
Fits the scenario-tree model (Q1) — three labeled scenarios with per-scenario evidentiary grade rather than a new output type.
🤖 Generated with Claude Code
From prod MCP feedback (2026-06-10, feedback_events #7): every sizing should output THREE time-anchored figures, each with its own evidentiary grade and expiry, never collapsed into one CAGR.
Guardrails (from the feedback's suggested_value JSON): report three figures separately with grades; no CAGR blending across grades; decisions must declare which figure they price against; units consistent per figure (GAAP vs ARR vs annual potential); separate expiries (run-rate expires ~quarterly in hypergrowth).
Worked example given: AI coding tools — realized 2025 ~$3-4B (B), run-rate 2026 ~$7-10B (B-), prize $50-150B (framework-only).
Fits the scenario-tree model (Q1) — three labeled scenarios with per-scenario evidentiary grade rather than a new output type.
🤖 Generated with Claude Code