I want to walk through a full engagement from start to finish
Not a highlight reel. The whole thing, including the parts that were slower and messier than I expected
B2B SaaS company. 12 employees. AWS account. Monthly spend when I started: $19,200
Their CTO had looked at the bill and estimated it should be around $8,000 to $10,000 for their workload. He was right. That instinct was correct. But he didn't know where the extra $9,000 was coming from and didn't have time to find it
Day one: I asked for read-only access to the account and access to their CloudWatch metrics. No write access at this stage. I wanted to understand the whole picture before touching anything
Day one findings: six areas worth investigating. CloudWatch Logs retention policies not set. RDS over-provisioned. EC2 instances with CPU utilization under 15 percent consistently. S3 Standard storage on objects not accessed in over a year. Data transfer between regions. One completely unknown service generating $800 a month that nobody could explain immediately
I documented each finding with the resource identifiers, current cost, and initial hypothesis for what was wrong
Day two: I started with the unknown service. Always start with the unknown. It turned out to be an ElasticSearch cluster (their term - AWS OpenSearch) that had been provisioned during a performance investigation, determined to not be needed, and never decommissioned. $800 a month for eight months: $6,400 total spent on something they'd decided not to use before it even launched
That discovery set the tone for the rest of the engagement. It told me this account had multiple things running that weren't being actively used. Which meant the real savings were larger than the obvious inefficiencies
We hadn't made a single change yet. We had a map
The map made everything that followed faster and more confident
Top comments (0)