When a global fashion retailer came to UP2CLOUD, their cloud bill had grown 140% in two years — yet their business had grown only 40%. The culprit wasn't rapid scaling. It was cloud sprawl: years of ad-hoc provisioning, forgotten environments, and an absence of governance. Within six weeks, we cut their monthly bill by 42%, saving $4,200 every month — and the changes kept compounding.
The Problem: Invisible Cloud Waste
Cloud waste is insidious precisely because it hides in plain sight. Most organisations don't have a single runaway workload — they have hundreds of small inefficiencies that collectively drain millions. In this client's case, a GCP + AWS audit revealed four categories of waste:
- Zombie resources — 37 compute instances that hadn't served traffic in over 90 days, including two full staging environments nobody had decommissioned after a product launch.
- Untagged workloads — 61% of resources had no cost-centre, environment, or owner tag, making it impossible to attribute spend or enforce budgets.
- Over-provisioned instances — Average CPU utilisation across production was 12%. Dozens of n1-standard-8 VMs were doing the work of n1-standard-2s.
- On-demand pricing for predictable workloads — Every long-running database and application server was paying on-demand rates despite running 24/7 for over a year.
The UP2CLOUD FinOps Audit Methodology
Our FinOps audit is a structured five-phase engagement designed to surface, quantify, and remediate waste — without disrupting running services. Here's how each phase played out.
Phase 1 — Inventory & Tagging Strategy
Before you can optimise, you need visibility. We deployed a tagging taxonomy across all GCP projects and AWS accounts: env (prod/staging/dev), team, cost-centre, managed-by (terraform/manual), and last-reviewed. Terraform modules were updated to enforce tags at creation time. For existing untagged resources, we ran a GCP Asset Inventory export and cross-referenced it with AWS Config to auto-classify resources by usage pattern.
Phase 2 — Rightsizing
Using GCP Recommender and AWS Compute Optimizer, we generated a rightsizing report for every instance family. We applied a conservative approach: only downsize instances where the P99 CPU utilisation over 30 days was below 30%. This eliminated risk while still capturing the bulk of waste. The result was a migration from 40 over-provisioned VM families to appropriately sized equivalents — reducing compute spend by 28% in isolation.
Phase 3 — Reserved Instances & Committed Use Discounts
After rightsizing, we could commit to steady-state capacity with confidence. We purchased 1-year Committed Use Discounts (CUDs) on GCP for the production database fleet and 1-year Reserved Instances on AWS for the application tier. This alone delivered a further 35% discount on the committed workloads. The principle: never buy commitments before rightsizing, or you lock in waste.
Phase 4 — Spot & Preemptible Fleets for Batch Workloads
The client ran nightly ETL jobs and ML training pipelines on on-demand instances. These were perfect candidates for GCP Preemptible VMs and AWS Spot Instances — workloads that tolerate interruption in exchange for up to 80% cost reduction. We wrapped them in a managed instance group with automatic restart logic, and the total cost of batch processing dropped by 74%.
Phase 5 — Continuous Governance with Terraform & Budget Alerts
Point-in-time savings decay without governance. We implemented GCP Budget Alerts at the project level (80% and 100% thresholds) and AWS Cost Anomaly Detection for every service. All infrastructure was migrated to Terraform with enforced tagging policies using OPA Conftest in the CI/CD pipeline — any resource missing mandatory tags fails the plan before it reaches apply.
Tools Used
- GCP Billing & Recommender — cost visibility, rightsizing recommendations
- AWS Cost Explorer & Compute Optimizer — spend analysis, instance recommendations
- Terraform + OPA Conftest — infrastructure as code, policy enforcement
- GCP Asset Inventory + AWS Config — resource discovery and compliance
- PagerDuty + Slack — budget alert routing and notifications
Results After 6 Weeks
The combined impact of all five phases delivered a 42% reduction in monthly cloud spend — from approximately $10,000/month to $5,800/month, a saving of $4,200 every month. Annualised, that's $50,400 returned to the business. More importantly, the client now has a governance framework that prevents the drift from recurring: tagged resources, committed budgets, enforced Terraform policies, and automated anomaly detection.
Cloud cost optimisation is not a one-time project. It's an ongoing discipline — FinOps. The organisations that treat it as such consistently outperform peers by maintaining lean cloud estates that scale efficiently as the business grows.