NorthStack
−38% cloud spend
NorthStack's AWS bill was growing faster than ARR. We stood up a multi-account landing zone, right-sized everything, and put FinOps guardrails in place — cutting spend 38% with zero downtime.
What was broken
NorthStack ran everything in a single AWS account with no tagging, oversized EKS nodes, and three idle RDS replicas. The bill had grown 4× in 18 months while revenue grew 1.8×. There were no budgets, no anomaly alerts, and SOC 2 auditors had flagged the lack of account separation. Two engineers were spending ~30% of their week on infra firefighting.
What we did
- 1
Designed a 4-account landing zone (prod / staging / shared-services / sandbox) using AWS Control Tower and Terraform modules.
- 2
Right-sized EKS node groups based on 90 days of CloudWatch + Datadog data. Moved batch workloads to Spot, dev to t-family on schedule.
- 3
Decommissioned two of three RDS replicas, moved analytics reads to a read-only replica with smaller instance class, enabled Performance Insights.
- 4
Set up Cost Anomaly Detection, per-team budgets, and a weekly FinOps digest that emails engineering leads. Every resource is now tagged owner+env+service.
- 5
Migrated 47 services account-by-account over 6 weekends. Zero customer-visible downtime; one 11-minute internal maintenance window.
How it unfolded
Cost + architecture audit. SOC 2 gap analysis. Roadmap signed off.
Control Tower, 4 accounts, Terraform modules, SSO.
47 services moved account-by-account. Zero customer downtime.
EKS, RDS, networking. Spot for batch. Reserved Instances re-purchased.
Budgets, anomaly alerts, weekly digest, runbook.
“They cut our bill 38% and gave us a cloud setup the auditors actually liked. Two of our engineers got their weeks back.”