Finops Analyst

FinOps Analyst

When to Use

• Analyze cloud bills and usage (CUR, Cost Explorer, billing export, Cost Management)
• Design tagging and allocation keys for showback or chargeback
• Find waste — idle compute, unattached storage, orphaned IPs, over-provisioned SKUs
• Recommend commitments — RIs, Savings Plans, CUDs, reservations (usage-based)
• Build budgets, forecasts, and variance explanations for leadership
• Detect anomalies and investigate spend spikes
• Report unit economics — cost per tenant, API, feature, or environment
• Run FinOps reviews with service owners (monthly optimization cadence)
• Model what-if scenarios (region move, architecture change, commit purchase)

When NOT to Use

• Map cloud spend to GL, amortize prepaids, fixed assets → compute-accounting-manager
• Enterprise agreement strategy and org-wide CCoE FinOps program → enterprise-cloud-architect
• Provision VPC, fix IAM, implement infra → cloud-engineer
• Execute access tickets and patch windows → cloud-system-administrator
• Rack/GPU utilization and stranded kW → data-center-compute-supply-efficiency
• Capex policy and board asset governance → director-infrastructure-capex-accounting
• Generic BI without cloud cost domain → bi-analyst
• SOC 2 cloud evidence → cloud-compliance-specialist
• TCO/NPV business cases, commit portfolio economics, CFO scenarios → cloud-economist
• Cloud program strategy, migration portfolio, EA sign-off → vp-of-cloud
• Infrastructure org strategy, portfolio budget, board narratives → vp-of-infrastructure

Related skills

Core Workflows

1. Scope and FinOps principles

Inform, optimize, operate; boundaries with accounting.

See `references/finops_scope_and_principles.md`.

2. Cost visibility and allocation

Tags, CUR, dashboards, allocation rules.

See `references/cost_visibility_allocation.md`.

3. Optimization and rightsizing

Waste, SKUs, commitments, engineering tickets.

See `references/optimization_rightsizing.md`.

4. Budgeting and forecasting

Budgets, forecasts, anomalies.

See `references/budgeting_forecasting.md`.

5. Unit economics and reporting

Cost per unit, executive narratives.

See `references/unit_economics_reporting.md`.

6. FinOps governance

Policies, RACI, review cadence.

See `references/finops_governance.md`.

Outputs

• Cost breakdown — by account, service, tag, trend vs prior period
• Optimization backlog — savings estimate, owner, risk, effort
• Commitment recommendation — coverage %, term, break-even, risk
• Allocation model — keys, untagged %, disputed splits
• Forecast — baseline, growth drivers, confidence range
• Anomaly report — root cause, one-time vs recurring

Principles

• Everyone owns their spend — FinOps enables; engineering decides trade-offs
• Unit economics over totals — tie cost to business drivers
• Measure then commit — stabilize usage before multi-year RIs
• Same tags as finance — align with compute-accounting-manager keys where possible
• Fast feedback — weekly visibility beats quarterly surprises

Budgeting and forecasting

Table of contents

1. Budgets 2. Forecast methods 3. Variance analysis 4. Anomaly detection

Budgets

Per account, BU, or tag value:

• Set monthly budget with 80% and 100% alert thresholds
• Separate prod vs non-prod budgets
• Include expected credits and EA drawdown if applicable
• Review budget vs actual in monthly FinOps meeting

Budget is a planning tool; GL forecast may differ (compute-accounting-manager).

Forecast methods

Present range (low/base/high) for leadership; document assumptions.

Variance analysis

Monthly narrative:

• Volume — usage growth (instances, requests, GB)
• Rate — price list, new services, cross-region egress
• One-time — data transfer spike, misconfiguration, pentest environment left on
• Commitment — amortization true-up, unused RI

Compare to forecast and prior month; flag >10% movers for investigation.

Anomaly detection

Enable native anomaly alerts (AWS, GCP, Azure) plus custom rules:

• Daily spend >2σ from 14d baseline per account
• New service first appearance in top 10
• Tag value sudden 3× increase

Triage within 48h: real growth, leak, or attack/abuse (loop security if suspicious).

Cost visibility and allocation

Table of contents

1. Data sources 2. Tagging standard 3. Allocation methods 4. Untagged spend

Data sources

Use daily granularity for ops; monthly for leadership. Include credits, refunds, and tax lines explicitly.

Tagging standard

Minimum tags (enforce via policy — cloud-security-engineer / cloud-engineer):

Activate cost allocation tags in billing console. Report untagged % weekly.

Allocation methods

Document method per shared platform (EKS, data lake, NAT gateway).

Untagged spend

1. Rank by $ untagged descending 2. Identify resource type (often EC2, RDS, S3 without owner) 3. Open ticket to owner with deadline to tag or decommission 4. Escalate repeat offenders to engineering management 5. Apply default allocation only as temporary with expiry

Target: <5% untagged spend at steady state (org-specific goal).

FinOps governance

Table of contents

1. RACI 2. Cadence 3. Policies 4. Engineering partnership

RACI

A = accountable, R = responsible, C = consulted, I = informed.

Cadence

Policies

Document and enforce:

• Sandbox auto-shutdown or budget cap
• Prod requires tags before deploy (policy-as-code)
• Commit approval threshold (e.g., >$50k annual)
• New region/service FinOps review for cost estimate
• Anomaly response SLA

Exceptions with expiry in register.

Engineering partnership

• Embed FinOps review in PRR (site-reliability-engineer) for cost estimate
• Celebrate verified savings in team metrics
• No punitive chargeback without visibility first
• Provide self-service dashboards before demanding cuts

Escalate chronic untagged spend to engineering director with data, not anecdotes.

FinOps scope and principles

Table of contents

1. FinOps phases 2. Role boundary 3. Shared metrics 4. Anti-patterns

FinOps phases

Run all three; do not skip Inform and jump to commitment purchases.

Role boundary

FinOps does not post journal entries or capitalize hardware.

Shared metrics

Align with finance where possible:

• Same tag keys (cost-center, environment, service)
• Same billing account hierarchy in reports
• Document lag (CUR 24h+, invoice monthly)

Reconciliation meeting: FinOps dashboard vs GL flux (compute-accounting-manager).

Anti-patterns

• Buying 3-year RIs before workload is stable
• Optimizing only at month-end after invoice shock
• Chargeback without tagging discipline (arbitrary splits)
• Savings that harm SLOs without SRE sign-off (site-reliability-engineer)
• Blaming finance for untagged sprawl — fix process with engineering leads

Optimization and rightsizing

Table of contents

1. Waste categories 2. Rightsizing process 3. Commitments 4. Architecture levers

Waste categories

Use native recommendation engines (Compute Optimizer, Advisor) as input, not sole truth.

Rightsizing process

1. Identify candidate from metrics (CPU, memory, network, IOPS) 2. Validate with owner — batch jobs, peak windows, launch coming 3. Schedule change in maintenance window if prod 4. Measure 7d post-change for latency/errors 5. Record savings in optimization log

Never rightsizing without owner ack on stateful prod.

Commitments

Decision flow:

1. 30–60 days stable hourly usage pattern 2. Calculate on-demand vs RI/SP/CUD break-even 3. Choose term (1y vs 3y) vs flexibility risk 4. Target coverage (e.g., 70–80% of steady baseline; not 100% on volatile) 5. Track utilization monthly; sell or exchange if provider allows

Enterprise EA and payer-level commits → enterprise-cloud-architect.

Architecture levers

Engage cloud-architect / service teams for larger wins:

• Graviton/ARM migration
• Serverless vs always-on
• S3 lifecycle and storage class
• Cross-AZ traffic reduction
• Caching to cut data egress
• Regional consolidation for residency + cost

Document reliability impact before architecture-driven cuts.

Unit economics and reporting

Table of contents

1. Unit definitions 2. Calculation patterns 3. Executive narrative 4. Dashboards

Unit definitions

Pick units aligned to business model:

Document shared cost allocation to units (platform tax).

Calculation patterns

unit_cost = (direct_tagged_spend + allocated_shared_spend) / unit_count

Rules:

• Use same period for numerator and denominator
• Exclude one-time spikes from unit trend or footnote them
• Gross margin view: unit revenue − unit infra cost (with finance)

For multi-tenant: allocate by usage meter not equal split unless fair.

Executive narrative

One-page monthly:

• Total cloud spend vs budget and vs last month
• Top 3 drivers of change
• Savings delivered (verified) vs backlog
• Commitment utilization %
• Untagged % and top offenders
• Forward look — launches, commits, risks

Avoid raw service lists without interpretation.

Dashboards

Layers:

1. Company — total, by payer account, budget burn 2. BU/product — tagged allocation 3. Service owner — drill to resource level 4. Optimization — open recommendations $

Self-serve for engineering; FinOps maintains data quality and definitions.