Implementing FinOps Practices for Cloud Cost Management

Your cloud resource bill grows faster than your business. Recently, a company approached us: over six months, their AWS bill grew from $20k to $80k. Our audit revealed 35% idle RDS and NAT Gateway. We implemented FinOps: set budgets, auto-shutdown dev environments, and purchased Savings Plans. After 8 weeks, the bill dropped by 40%. FinOps is an operational model for managing costs that unifies engineers, finance, and business. Without such practices, companies lose up to 40% of their cloud budget.

How FinOps Practices Manage Cloud Costs?

FinOps is not just a set of tools—it's a culture of shared responsibility. We implement it from scratch or refine existing processes. Our approach: quick wins immediately (rightsizing, removing abandoned resources) and a long-term strategy: commitment, automation, culture. The result: a 20–40% reduction in bills without performance loss. According to the FinOps Foundation, maturity is divided into three levels: Crawl, Walk, Run.

How FinOps Identifies Hidden Losses?

Three sources of losses found in 80% of projects:

• Idle and underutilized resources. EC2 instances with 3–5% CPU utilization running 24/7. RDS with Multi-AZ in dev environments. NAT Gateway in a region with no traffic. All easily detectable via AWS Trusted Advisor or Cost Anomaly Detection, but without regular monitoring they remain unnoticed.
• Lack of Savings Plans and Reserved Instances. On-demand pricing is the highest rate. Compute Savings Plans with a 1-year commitment give 20–40% discounts on EC2, Lambda, and Fargate. Many teams pay on-demand for years simply because they "haven't figured it out" or are afraid to commit.
• Data transfer costs. Egress between regions, between AZs, through NAT Gateway—these lines in the CUR (Cost and Usage Report) are often ignored until they take up 15–20% of the bill. Typical case: microservices in different AZs pushing traffic through NAT instead of VPC Endpoint or PrivateLink.

How We Implement FinOps: Three Maturity Levels

The FinOps Foundation divides maturity into Crawl → Walk → Run. In practice:

Crawl — Visibility. Set up Cost and Usage Reports with hourly granularity, export to S3, create an Athena table on top. Mandatory tags on all resources (project, team, environment). Weekly cost review meetings. At this level, most companies already find 15–25% potential savings, which in monetary terms can be $10,000–$50,000 per month on a $50,000–$200,000 bill.

Walk — Optimization. Rightsizing instances based on CloudWatch metrics (CPU/Memory over 30 days). Migrating to Graviton3 where the stack allows. Purchasing Savings Plans for baseline load, on-demand only for peaks. Automatic shutdown of non-production environments on schedule (EventBridge + Lambda or Instance Scheduler).

Run — Automation and Predictability. Budgets with alerts in AWS Budgets. Cost Anomaly Detection with SNS notifications. FinOps metrics on engineering dashboards alongside SLOs. Chargeback/showback per team via Cost Allocation Tags.

Why FinOps Implementation Requires Cultural Changes?

Even after implementing basic practices, growth can continue due to lack of a culture of shared responsibility. We set up processes so that each team sees their costs and can influence them. A weekly 30-minute cost review with representatives from different teams is the minimum format to maintain culture. The unit economics metric—cost per transaction, per user, per request—shows optimization effectiveness better than absolute bill numbers.

How Automation Reduces Costs?

Instance Scheduler

First thing we implement in dev/staging—automatic shutdown of unused resources:

---
# CloudFormation parameter for Instance Scheduler Periods: office-hours: begintime: "08:00" endtime: "20:00" weekdays: "mon-fri" timezone: "Europe/Moscow" 
---

Cost Anomaly Detection

For automatic anomaly detection, set up via AWS CLI:

aws ce create-anomaly-monitor \
    --anomaly-monitor '{"MonitorName":"TeamCostMonitor","MonitorType":"DIMENSIONAL","MonitorDimension":"SERVICE"}' \
    --region us-east-1

aws ce create-anomaly-subscription \
    --anomaly-subscription '{
        "SubscriptionName": "DailyCostAlert",
        "MonitorArnList": ["arn:aws:ce::123456789012:anomalymonitor/..."],
        "Subscribers": [{"Address": "[email protected]", "Type": "EMAIL"}],
        "Threshold": 20,
        "Frequency": "DAILY"
    }'

Step-by-step instructions:

1. Create an anomaly monitor using aws ce create-anomaly-monitor.
2. Set up a notification subscription via aws ce create-anomaly-subscription.
3. Specify a threshold (Threshold)—for example, 20% deviation from expected.
4. Test by triggering a test anomaly or wait for a real one.

Rightsizing

Look at CloudWatch metrics over 30 days: if CPUUtilization p95 is below 20%, it's a candidate for downsizing. For example, m5.2xlarge → m5.xlarge saves about $70/month per instance. Multiply by 50 instances—that's a substantial amount.

Savings Plans: How to Calculate Commitment

A common mistake is buying Savings Plans for peak load. The right approach is to cover only baseline, leaving the rest on-demand. From Cost Explorer, take hourly spend over the last 3 months, find the minimum value—that's the safe commitment. For example, if the minimum hourly spend is $8, commitment = $8/hour on Compute Savings Plan. Beyond that, use on-demand or Spot for batch jobs. This calculation avoids overpaying for commitment.

What's Included in Our Work

Stages and Timelines

Full cycle from audit to working processes—6–10 weeks for a company with 5–15 engineers and one cloud account. For multi-account structures with multiple teams—3–4 months. Our experience: 5+ years in DevOps, 50+ cloud cost optimization projects, AWS certified engineers. Request a FinOps audit to find out how much you can save. Contact us for a consultation.