· Cloud
E-commerce Platform
Problem
AWS spend had grown 3× year-on-year with no cost visibility and no one owning the bill.
Outcome
38% cost reduction in 60 days, zero performance regression, full tagging and attribution in place.
Key metrics
- Cost reduction
- 38%
- Time to result
- 60 days
- Performance regression
- None
What we did
The client's AWS bill had tripled in 18 months but nobody could say where the money was going. Resources were untagged, environments bled into each other, and the engineering team had no cost tooling in their daily workflow.
We started with a tagging audit and built a cost attribution model — splitting by environment, service, and team. Within a week the client had visibility they'd never had. The first round of quick wins (right-sizing over-provisioned RDS instances and retiring zombie resources) cut 12% off the bill immediately.
The deeper work was migrating stateless compute to Graviton instances and moving batch workloads to Spot. We ran both changes in staging first, validated throughput under synthetic load, and promoted to production with feature flags.
At the 60-day mark the bill was down 38% from baseline with no change in p95 latency. We left behind a Terraform module for enforcing tagging on all new resources and a monthly cost review runbook that the engineering lead now runs independently.
Have a similar problem?
Start a project →