Forecasting fits a curve to historical data and extrapolates it forward. Excel and tools like fityk report an R² (coefficient of determination) — how well the equation matches the data; 1.0 is perfect, and for capacity work anything above ~0.85 is usually sufficient. As R² drops, so should your confidence. But a high R² is a trap. Two rules keep curve-fitting honest: - **Avoid high-order polynomials.** Anything above 2nd order fits the known data beautifully yet swings wildly *outside* it — exactly where your forecast lives. - **Context beats fit.** A 99.999%-accurate curve is still wrong if it ignores the data's real-world drivers (you can't forecast snow-shovel sales without season and geography). A near-perfect fit is often the product of questionable assumptions. Use common sense; don't insist on elegant fits. The goal is a defensible educated guess, not mathematical beauty. --- *Source: [[The Art of Capacity Planning]] (John Allspaw, O'Reilly 2008) — Ch 4 — Predicting Trends*