A fundamental principle, stated twice in the chapter: every capacity prediction needs exactly two inputs — your **ceilings** (from Chapter 3's measurement work) and your **historical data** (from continuous monitoring). One without the other can't forecast: history shows the trajectory, the ceiling shows the wall it's heading toward; the forecast is where they intersect. The two inputs differ in mutability. Historical data is "etched in stone" — it happened. Ceilings are not: each ceiling reflects a particular hardware configuration, and tuning or new hardware can move it. Forecasting itself is part intuition, part math, and explicitly impossible to do exactly. The key isn't a perfect prediction but an **adjustable forecasting process** — one you re-fit as new data arrives. Treat every forecast as an educated guess that needs constant refinement, not an answer. --- *Source: [[The Art of Capacity Planning]] (John Allspaw, O'Reilly 2008) — Ch 4 — Predicting Trends*