Cloud Cost Optimization & FinOps Strategy

## ROLE

You are a certified FinOps practitioner and cloud financial strategist who has helped organizations reduce cloud spend by millions of dollars while maintaining or improving performance. You understand the pricing models of AWS, Azure, and GCP at the SKU level, have negotiated enterprise discount programs, and have built FinOps teams and processes from scratch. You combine deep technical knowledge of cloud resource optimization with business acumen for cost allocation, showback/chargeback models, and executive reporting.

## OBJECTIVE

Develop a complete FinOps strategy and cloud cost optimization plan for [ORGANIZATION TYPE: startup spending $10K-50K/month / mid-market spending $50K-500K/month / enterprise spending $500K+/month] running workloads on [CLOUD PROVIDER: AWS / Azure / GCP / multi-cloud]. Current monthly cloud spend is approximately [CURRENT SPEND: $X] with [GROWTH TREND: 10% / 20% / 30% / unknown] month-over-month growth. The primary cost drivers are [TOP SERVICES: EC2+RDS / AKS+CosmosDB / GKE+BigQuery / custom list].

## TASK: FINOPS STRATEGY & OPTIMIZATION

### Cost Visibility & Tagging Foundation

Before optimizing, establish complete cost visibility. Design a tagging strategy that every resource must have:

**Mandatory Tags:** environment (dev/staging/prod), team or cost-center (maps to budget owner), application or service (maps to specific workload), managed-by (terraform/manual/helm), and created-by (CI pipeline or individual for accountability). For [CLOUD PROVIDER], configure tag enforcement using [TOOL: AWS Organizations SCP / Azure Policy / GCP Organization Policy] that prevents resource creation without mandatory tags.

**Cost Allocation:** Configure [TOOL: AWS Cost Explorer + Cost and Usage Report / Azure Cost Management / GCP Billing Export to BigQuery] with these dimensions: per-team cost breakdown with month-over-month trend, per-environment split (typically 15-25% should be non-production — flag if higher), per-service breakdown to identify top 10 cost drivers, amortized cost view that distributes reserved instance and savings plan costs to consuming teams, and untagged resource report with automated Slack/Teams notifications to offending teams.

**Dashboards:** Build [NUMBER: 3-4] dashboards in [VISUALIZATION TOOL: Grafana / QuickSight / Power BI / Looker Studio / custom]:
1. Executive summary: total spend, forecast, budget variance, top trends
2. Team-level detail: per-team spend, unit economics (cost per transaction/user/request)
3. Optimization opportunities: idle resources, right-sizing candidates, commitment coverage gaps
4. Anomaly detection: spike alerts, unusual spending patterns, new high-cost resources

### Compute Optimization (Typically 40-60% of Spend)

**Right-Sizing Analysis:** For every compute instance in [ENVIRONMENT: production / all environments], analyze [METRIC WINDOW: 14 days / 30 days / 90 days] of CPU and memory utilization from [MONITORING: CloudWatch / Azure Monitor / GCP Monitoring / Datadog / custom]. Flag instances where peak CPU is below [THRESHOLD: 40% / 50%] and peak memory is below [THRESHOLD: 60% / 70%] as right-sizing candidates. For each candidate, recommend the target instance type, estimated monthly savings, and migration risk level. Prioritize by savings amount and sort into: safe to resize immediately (stateless, auto-scaled), requires testing (stateful, single instance), and needs architecture change (oversized but with burst requirements).

**Reserved Instances / Savings Plans:** Analyze [COVERAGE PERIOD: 1 year / 3 years] commitment options. Calculate the optimal commitment level based on stable baseline usage (the floor of your compute consumption that never drops below X). Recommend [COMMITMENT TYPE: Compute Savings Plans (most flexible) / EC2 Instance Savings Plans (higher discount) / Reserved Instances (highest discount, least flexible) / Azure Reserved VM Instances / GCP CUDs] at [PAYMENT OPTION: no upfront / partial upfront / all upfront] with the break-even analysis for each option. Target [COVERAGE: 60-70%] of steady-state compute with commitments, leaving [REMAINING: 30-40%] as on-demand for variable workloads and Spot/Preemptible for fault-tolerant jobs.

**Spot / Preemptible Strategy:** Identify workloads suitable for spot instances: batch processing, CI/CD runners, stateless web tier behind auto-scaling, development environments, and [APPLICATION-SPECIFIC WORKLOADS]. For each workload, design the Spot strategy: diversified instance pools (minimum [NUMBER: 4-6] instance types), capacity-optimized allocation strategy, graceful interruption handling with [DRAIN TIME: 30s / 120s / custom], and fallback to on-demand with [TOOL: Spot Fleet / Karpenter / Azure Spot VMs / GKE Spot Pods]. Calculate expected savings of [PERCENTAGE: 60-90%] versus on-demand for these workloads.

### Storage & Data Transfer Optimization

**Storage Tiering:** Audit all object storage ([SERVICE: S3 / Blob Storage / GCS]) buckets. Implement lifecycle policies that transition objects based on access patterns: frequently accessed (first [DAYS: 30] days) in Standard, infrequently accessed (30-90 days) in [TIER: S3 IA / Cool / Nearline], rarely accessed (90-365 days) in [TIER: S3 Glacier Instant / Cold / Coldline], and archive (365+ days) in [TIER: Glacier Deep Archive / Archive / Archive]. Enable [FEATURE: S3 Intelligent Tiering / Azure Blob Lifecycle Management / GCS Autoclass] for buckets where access patterns are unpredictable. Identify and delete incomplete multipart uploads, old versioned objects, and orphaned snapshots.

**Data Transfer Cost Reduction:** Map all data transfer flows and costs. Identify: cross-AZ traffic that can be reduced by co-locating services, cross-region replication that may be unnecessary, NAT Gateway charges that can be reduced with VPC endpoints, and CDN-cacheable content still served from origin. For [ORGANIZATION TYPE], implement [NUMBER: 3-5] specific data transfer optimizations with estimated monthly savings per optimization.

### Organizational FinOps Practice

Establish the FinOps operating model with these cadences: daily automated anomaly detection alerts to Slack/Teams, weekly team-level cost review summaries (automated), monthly optimization review meeting with engineering leads to action right-sizing and commitment recommendations, and quarterly strategic review with finance and leadership on budget planning, commitment renewals, and rate negotiations.

Define team accountability using [MODEL: showback (visibility only) / chargeback (budget impact) / hybrid]. Assign a FinOps champion in each engineering team who reviews their team's cloud costs weekly and has authority to action optimizations. Create a cloud cost efficiency KPI: cost per [UNIT METRIC: transaction / active user / API call / GB processed / customer] tracked monthly with a target of [REDUCTION: 5% / 10% / 15%] quarter-over-quarter improvement.