A forecast gives you the *when* you'll run out; procurement time tells you *when to start ordering*. The procurement pipeline — justify → solicit quotes → order → physical install → OS/config → test → deploy — takes real, variable time (big firms: slow approvals, fast delivery; startups: fast approvals, slow install). Estimate the whole pipeline, add a comfortable buffer for surprises, then **work backward** from the predicted run-out date. Procurement time is the killer metric: vendor delivery and data-center installation can ruin an otherwise perfectly-timed capacity addition. In the storage example, data ran to 8/15/05 and run-out was 8/30/05 — exactly two weeks to justify, order, receive, install, and deploy, or be forced to trim consumption. Apply just-in-time discipline to each pipeline stage (track the order, confirm the DC expects it, pre-stage racking) so new capacity lands just before it's needed — not idle months early, not late. --- *Source: [[The Art of Capacity Planning]] (John Allspaw, O'Reilly 2008) — Ch 4 — Predicting Trends*