Cloud cost when money stops being free
Three clients asked us for the same thing in one month: cut the bill without touching the product. It can be done, and it's boring.
Sixty percent of the savings came from test environments left running overnight and snapshots nobody deleted. A simple shutdown schedule on test stacks recovered more budget than any code optimization would have. The low-hanging fruit was always there, just invisible in the monthly bill.
The test account was particularly wasteful: development databases spinning at production costs even during off hours. Tagging resources by environment and automating overnight termination reclaimed thousands monthly. This pattern repeated across three of the four clients.
The next twenty-five percent was right-sizing, not optimizing code. Instances launched with default sizing remained unchanged years later, even as usage fell. Measuring actual peak load and matching instance types to real demand cost nothing to discover and everything to implement before the audit.
Reserved instances and savings plans offered deep discounts for committed spend, but only if you had confidence in your baseline. We ran the retrospective query on six months of metrics to find the floor before recommending commitment.
Only the last slice needed architectural change, and that's where we stopped: the rest wasn't worth the risk. A major redesign might have cut another ten percent, but the disruption and regression risk outweighed a modest savings.