Cost Model Sensitivity
Identifying which assumptions in a cost projection dominate the outcome, so effort goes into the estimates that actually matter.
Cost models are built with a dozen assumptions and are usually dominated by two or three. Sensitivity analysis finds which, and it changes both where the estimating effort goes and where the risk is.
The method is simple: vary each assumption independently across a plausible range and observe the effect on the total. An assumption that moves the total by 40% deserves research; one that moves it by 2% deserves a guess and a footnote.
The assumptions that typically dominate: request or user volume, since most costs scale with it; data volume and growth rate, which drive storage and transfer and compound; cache hit rate, because a fall from 99% to 95% multiplies backend load by five; cross-region or cross-zone traffic volume, which is frequently under-estimated by an order of magnitude; and the utilisation target, which sets how much capacity must exist for a given load.
Two disciplines that keep a model useful rather than decorative:
Model a range, not a point. Present low, expected and high scenarios, because planning to the central estimate means being wrong half the time and having no prepared response.
Compare the model to actuals after deployment. A model never checked against reality trains the organisation to distrust all of them; one that is checked improves quickly.