All work
B2B SaaS · SaaS platform (anonymized)

Cut AWS bill ~42% without touching reliability

Rebuilt the infrastructure around right-sizing and auto-scaling. Same uptime, dramatically lower monthly spend.

  • AWS
  • DevOps
  • Cost optimization
−42%Monthly AWS spend
99.98%Uptime maintained
< 4 wksAudit → first savings

Timeline: ~4 weeks · Audit → right-size → migrate

The problem

Monthly AWS spend was growing every month while traffic was flat — a classic sign of oversized instances and orphaned resources.

The team had layered features faster than they had revisited infra, leaving production running on reservations and instance types from 18 months earlier.

Every cost review ended in "we should look at this someday" because nobody wanted to risk reliability.

Our approach

  • Full account audit with Cost Explorer + tagged-resource breakdown. Found the top 5 cost drivers in week one.
  • Right-sized RDS and EC2 based on real CloudWatch percentiles, not guesswork. Migrated to Graviton where workloads supported it.
  • Replaced always-on workers with auto-scaling groups + scheduled scale-to-zero on non-prod environments overnight.
  • Moved batch and analytics workloads to spot capacity with proper interrupt handling.
  • Set up budget alerts and a monthly cost-review dashboard so the savings stick.

The result

AWS spend dropped ~42% in the first full month after cutover and stayed there.

Zero customer-visible incidents during the migration — every change was rolled out behind feature flags or via blue/green.

The cost dashboard is now part of the team's monthly ops review, so cost regressions get caught the same week they happen.

Our AWS bill was climbing every month with traffic we did not actually have. NEROOM rebuilt our infra around auto-scaling and right-sizing — same reliability, dramatically lower bill.
CTO, SaaS platform (anonymized)

Stack

  • AWS (EC2, RDS, Lambda, S3)
  • Terraform
  • CloudWatch
  • Spot instances
  • Auto Scaling Groups

Like to talk about something similar?

We only take on a small number of new projects each month. If your team has a problem in this shape — or something completely different — start with a free 30-minute call.

More work