If you Google “cloud cost optimization,” you will get a hundred articles telling you to buy reserved instances and rightsize your VMs. That advice is correct as far as it goes — but for most growing SaaS companies, it captures maybe 20% of the available savings.
This post covers the framework we actually use for cloud cost work, including the categories of waste most teams never look at.
Start with attribution, not optimization
Before you can optimize, you need to know where the money is going. Most teams have one number — “the AWS bill went up 15% this month” — and no idea which service, team, or environment drove it.
The first move is always tagging discipline:
Without this, every cost conversation is opinion-based.
The waste categories most teams miss
Idle and forgotten resources
Old EC2 instances from a long-completed POC. Load balancers attached to nothing. Elastic IPs not associated with running resources. Snapshots from 2023. Unused EBS volumes. These add up faster than people expect — we routinely find 5–12% of monthly spend in zombie resources.
Over-provisioned managed services
RDS instances at 5% CPU. ElastiCache clusters with 10% memory utilization. NAT Gateways processing 20MB/day at $32/month each. Managed services are convenient and that’s exactly why teams stop questioning their sizing.
Cross-AZ data transfer
The silent killer. If you have services in multiple availability zones talking to each other, you’re paying $0.01/GB for that traffic. At scale, this can be a five-figure line item. Often a single-AZ deployment is fine for non-critical workloads.
Logging and observability spend
Datadog, New Relic, and CloudWatch costs scale with engineering team size and rarely get pruned. Old log groups, high-cardinality metrics, debug logging left on in production — all real money.
S3 storage tiers
Standard storage for data nobody has touched in 18 months. Intelligent Tiering is one toggle that pays for itself in weeks for most teams.
Egress traffic
Cloud providers charge for traffic leaving the network. CDN configuration, image optimization, and architectural choices about where data lives all affect this.
Non-production environments running 24/7
Dev and staging clusters running at full size on weekends and overnight. Auto-shutdown schedules can cut these costs in half with no engineering impact.
The framework we use
For Cloudvorn cloud cost engagements, we work in this order:
Done in this order, most growing SaaS companies see 20–40% cost reduction within the first quarter — sometimes more.
What this looks like as an engagement
Our [Cloud Cost & Performance Review](/services#cloud-cost) is a 1–2 week fixed-price engagement that runs this framework end-to-end and delivers a prioritized roadmap with projected savings. Pricing is fixed at $5,000 or 20% of validated first-year savings — whichever you prefer.
The takeaway
Cloud cost optimization is not just rightsizing. The biggest savings usually come from attribution, killing zombies, and architectural choices most teams never revisit. A structured, data-driven pass through your cloud bill almost always finds 20%+ to cut — without changing how your application behaves.