Eight months and a full infrastructure budget spent building the wrong architecture — because nobody asked which parts of the pipeline actually needed to be on-prem. Here's a component-by-component framework for deciding on-prem, cloud, or hybrid, based on what each workload actually requires.