Case Study
A growth-stage SaaS company had a cloud bill that had grown with the business but never been rationalized. A systematic infrastructure audit found $20,000 a month in recoverable spend hiding in plain sight.
Cloud infrastructure that grows organically is almost always optimized for the moment, not the whole. Each decision made sense when it was made — spin up a new VPC for a new environment, enable detailed monitoring to debug a problem, add redundancy to handle a specific incident. Nobody is wrong. But nobody is ever looking at the sum total and asking whether it still makes sense.
That gap is where the waste lives. Not in any one decision, but in the accumulation of decisions that were never revisited.
This client had 25 VPCs, most running three NAT Gateways each — including non-production environments where that level of redundancy served no real purpose. The cost was real. The justification had long since expired.
More surprising was the AWS Config bill. At $4,000 a month, it was one of the larger line items on the account. The cause: the company relied heavily on spot instances, which by design spin up and terminate constantly. AWS Config was recording every one of those events. Thousands of them. The volume was enormous. The operational value was effectively zero.
CloudWatch metrics told a similar story. Detailed, hourly data collection across workloads where daily resolution provided the same operational visibility. More data, same insight, higher cost.
Non-production environments were running full production-grade redundancy with no business requirement to justify it.
Every spot instance launch and termination was being recorded and charged for, generating enormous volume with effectively zero operational value.
Detailed collection intervals across workloads where the higher resolution added cost but no operational value.
NAT Gateway consolidation was straightforward. Non-production environments were reduced to configurations appropriate for their actual risk profile, not carried over from production defaults.
For AWS Config, we disabled recording for the specific EC2 resource types generating the churn. The Config bill dropped from $4,000 a month to under $10. The change took an afternoon. The compliance and security configuration recording that actually mattered was untouched.
CloudWatch metric collection intervals were adjusted on workloads where the data supported it. Operational visibility remained intact. The team could still see everything they needed to see — just without paying for hourly granularity on systems where nothing meaningful changes hour to hour.
Beyond the infrastructure changes, we helped the client take advantage of AWS discounting and flexible commitment options, reducing compute costs without locking the business into infrastructure it might outgrow.
Monthly AWS spend went from $45,000 to $25,000. $20,000 a month. $240,000 annualized. The engineering team did not change a single line of application code. The infrastructure did not change in any way the end user would notice. The monitoring coverage the team relied on remained intact.
The savings were not found in one place. They were found in five or six places, each modest on its own, collectively significant. That is almost always how it works. There is rarely one catastrophic mistake. There is usually a long list of reasonable decisions that were never revisited.
That is exactly the kind of audit an internal team rarely has time to do for itself.