Every cloud cost programme starts the same way: a dashboard, a list of oversized instances, and a satisfying one-off reduction. Six months later spend is back where it was, because nothing about how the workload consumes resources changed.
Durable cost control comes from architecture. Storage tiering matched to real access patterns, data transfer designed rather than incidental, elasticity that actually scales to zero, and enough tagging discipline that a team can see the cost of its own decisions in time to make a different one.