Cloud Cost Data and Analytics¶
📖 FinOps Capabilities - Cost Allocation - Cost allocation documentation
Overview¶
This document covers cloud cost data, analytics, allocation, and reporting. Understanding how to collect, analyze, and report on cloud cost data is a core FinOps engineering skill tested on the Certified Engineer exam.
Cost and Usage Data¶
Cloud Provider Billing Data¶
| Provider | Billing Data Source | Format | Granularity |
|---|---|---|---|
| AWS | Cost and Usage Report (CUR) | CSV/Parquet in S3 | Hourly/Daily |
| Azure | Cost Management exports | CSV in Blob Storage | Daily |
| GCP | Billing export to BigQuery | BigQuery tables | Hourly/Daily |
AWS Cost and Usage Report (CUR)¶
- Most detailed billing data available from AWS
- Line-item level detail for every charge
- Includes resource IDs, tags, usage type, pricing
- Exported to S3 in CSV or Parquet format
- Can be queried with Athena, Redshift, or QuickSight
- Updated multiple times per day
📖 AWS CUR - CUR documentation
Azure Cost Management¶
- Export cost data to Azure Blob Storage
- Integration with Power BI for visualization
- API access for programmatic cost retrieval
- Supports management group, subscription, and resource group scopes
- Amortized and actual cost views
📖 Azure Cost Management - Azure billing docs
GCP Cloud Billing¶
- Standard export to BigQuery (daily aggregation)
- Detailed export to BigQuery (resource-level, hourly)
- Pricing export to BigQuery (list prices)
- SQL queries for custom cost analysis
- Integration with Data Studio/Looker for dashboards
📖 GCP Billing Export - BigQuery billing export
Cost Allocation¶
Tagging Strategy¶
- Tags are the foundation of cost allocation
- Consistent tagging enables accurate showback/chargeback
- Minimum tags: team, environment, project, cost-center
- Enforce tagging via policies (AWS Config, Azure Policy, GCP Org Policy)
Tagging Enforcement¶
# AWS: Tag policy via Organizations
# Azure: Azure Policy for required tags
# GCP: Organization policy constraints
# Example enforcement levels:
1. Advisory: Report untagged resources
2. Preventive: Block resource creation without required tags
3. Corrective: Auto-tag resources based on rules
Shared Costs¶
| Cost Type | Allocation Method | Example |
|---|---|---|
| Shared infrastructure | Proportional split | Kubernetes cluster, networking |
| Support and premium | Even split or proportional | Enterprise support plans |
| Reserved capacity | Applied to matched usage | RI/SP coverage |
| Marketplace | Direct allocation | Third-party licenses |
| Data transfer | Proportional to usage | Cross-region data movement |
Showback vs Chargeback¶
| Model | Description | Accountability |
|---|---|---|
| Showback | Report costs to teams (informational) | Low - awareness only |
| Chargeback | Bill costs to team budgets | High - financial accountability |
| Hybrid | Showback with chargeback for overages | Medium - balanced approach |
- Start with showback - build awareness before charging
- Chargeback requires mature tagging and allocation
- Shared costs are the hardest to allocate fairly
Cost Analytics¶
Cost Dimensions¶
| Dimension | Use Case | Example |
|---|---|---|
| Service | Which cloud services cost most | EC2, RDS, S3 |
| Account/Subscription | Cost by business unit | prod-account, dev-account |
| Region | Geographic cost distribution | us-east-1, eu-west-1 |
| Tag | Cost by team, project, environment | team:platform, env:prod |
| Resource | Individual resource cost | specific EC2 instance ID |
| Time | Cost trends and anomalies | Daily, weekly, monthly |
Anomaly Detection¶
- Sudden cost spikes compared to historical baseline
- Unusual resource provisioning patterns
- Unexpected data transfer charges
- New services appearing in billing
- Cost changes not correlated with known deployments
Cost Forecasting¶
| Method | Description | Accuracy |
|---|---|---|
| Linear trend | Extend current trend line | Low (does not account for events) |
| Moving average | Average of recent periods | Medium |
| Seasonal decomposition | Account for repeating patterns | Medium-High |
| ML-based | Machine learning on historical data | High (with sufficient data) |
Key Reports¶
- Executive summary - Total spend, month-over-month change, top services
- Team breakdown - Cost per team with trend and budget comparison
- Waste report - Idle resources, unused capacity, optimization opportunities
- Commitment utilization - RI/SP usage rate and coverage
- Unit economics - Cost per transaction, cost per customer
Unit Economics¶
Definition¶
Unit economics connects cloud costs to business metrics, enabling teams to understand the cost of delivering business value.
Examples¶
| Business Metric | Cloud Cost Metric | Unit Cost |
|---|---|---|
| API requests | Compute + networking cost | Cost per 1M requests |
| Active users | Total infrastructure cost | Cost per active user |
| Data processed | Compute + storage cost | Cost per TB processed |
| Orders placed | End-to-end infra cost | Cost per order |
Calculating Unit Costs¶
Unit Cost = Total Cloud Cost for Service / Number of Business Units
Example:
- Monthly compute cost for order processing: $15,000
- Monthly orders processed: 500,000
- Cost per order: $15,000 / 500,000 = $0.03 per order
Track over time:
- Month 1: $0.03/order
- Month 2: $0.028/order (improved efficiency)
- Month 3: $0.035/order (investigate increase)
Benefits of Unit Economics¶
- Decouple growth from cost conversation (cost went up because orders went up)
- Enable meaningful comparisons across time periods
- Identify efficiency improvements vs simple cost increases
- Support business case for optimization investments
FOCUS (FinOps Open Cost and Usage Specification)¶
Overview¶
- Open standard for cloud cost and usage data
- Normalizes billing data across cloud providers
- Common schema for multi-cloud cost analysis
- Supported by FinOps Foundation and major cloud providers
📖 FOCUS Spec - FOCUS specification
Key FOCUS Columns¶
| Column | Description | Example |
|---|---|---|
| BillingAccountId | Billing account identifier | 123456789012 |
| BillingPeriodStart | Start of billing period | 2024-01-01 |
| ServiceName | Cloud service name | Amazon EC2, Azure VMs |
| ResourceId | Unique resource identifier | i-1234567890abcdef0 |
| UsageQuantity | Amount of usage | 730 hours |
| BilledCost | Amount billed | $52.56 |
| EffectiveCost | Cost after discounts/credits | $36.79 |
| Tags | Resource tags | {"team": "platform"} |
Building Cost Dashboards¶
Dashboard Design Principles¶
- Start with the question - What decision does this data support?
- Layer detail - Executive summary to drill-down detail
- Highlight anomalies - Call attention to unexpected changes
- Include context - Show budget, forecast, and historical comparison
- Enable action - Link to optimization recommendations
Dashboard Hierarchy¶
Level 1: Executive (total spend, trend, top 5 services)
Level 2: Team/Business Unit (per-team cost, budget variance)
Level 3: Service (per-service cost, utilization)
Level 4: Resource (individual resource cost and metrics)