Apptio Cloudability cheat sheet

FinOps is an operating model: Inform → Optimize → Operate, repeated continuously as a loop, not a one-time project.

Unit economics formula: unit cost = total cloud spend ÷ business unit count (transactions, customers, active users).

Showback reports cost to the owning team without charging it; chargeback posts actual cost to that team's budget or P&L.

Amortization spreads a commitment's upfront or monthly fee evenly across the usage period so each month carries its true share.

Unblended cost reflects the rate actually charged at the time of usage; amortized cost adds the commitment fee effect and is preferred for internal accountability.

Fully loaded cloud cost = infrastructure + support + licensing + shared/platform overhead allocated to the consumer.

Forecast accuracy is measured as variance between forecast and actual; track it monthly and report the trend, not a single month.

Allocation completeness = cost tagged or mapped to an owner ÷ total cost; untagged spend is a governance gap, not a rounding error.

Budget variance % = (actual − budget) ÷ budget; positive means overspend against plan.

Rightsizing: compare utilization against the next smaller instance family; act when sustained utilization sits well below the size threshold over a representative window.

Idle resource detection: unattached storage volumes, unused elastic IPs, stopped instances still billing for attached storage, and orphaned snapshots.

Savings Plans and Reserved Instances trade flexibility for discount; longer term and higher upfront payment yield deeper discounts but less coverage flexibility.

Commitment coverage = spend covered by commitments ÷ eligible spend; aim to cover the steady-state baseline, not peak or spiky usage.

Commitment utilization = commitment actually consumed ÷ commitment purchased; low utilization means over-purchase, high coverage with low utilization is a red flag.

Spot/preemptible capacity offers the deepest discount but is interruptible; suitable for fault-tolerant batch, not stateful production.

Storage lifecycle policies tier cold data to cheaper classes; the cost of retrieval and minimum-duration charges must be weighed against the savings.

Anomaly detection flags cost spikes against a learned baseline; the runbook is triage, owner notification, root cause, then remediation.

Showback of optimization opportunities should include estimated monthly savings and effort, so teams prioritize by savings per unit of effort.

Tagging policy defines mandatory keys (owner, cost center, environment, application); enforcement via policy-as-code or budget alerts on untagged spend.

Account and subscription structure is the primary allocation boundary; organizational hierarchy should mirror the business hierarchy.

A budget alert threshold (e.g., at 80% and 100% of forecast) triggers owner action; alerts without an owner and a runbook are noise.

Guardrails: service quotas, budget actions that stop or notify, and approval gates for new spend, applied at the platform or account level.

Chargeback policy must define shared-cost allocation method (proportional to usage, equal split, or fixed percentage) and be documented and agreed.

Cost allocation tags have a propagation delay; new tags may take up to 24 hours to appear in billing data.

A center of excellence or FinOps guild owns standards, reviews exceptions, and publishes a monthly cost review cadence.

Policy exceptions require an owner, a justification, and an expiry date; open-ended exceptions become permanent leakage.

Gross margin per unit = (revenue per unit − cloud cost per unit) ÷ revenue per unit; track the trend, not just the level.

Cost per transaction = total cloud cost attributable to the workload ÷ number of transactions in the period.

Cost per active user = total cloud cost ÷ monthly active users; segment by cohort to avoid averaging away the signal.

Elasticity: as volume grows, fixed platform cost is spread over more units, so unit cost should fall; rising unit cost signals inefficiency.

Attribution: direct costs map cleanly to a unit; shared costs must be allocated by a driver (requests, CPU-hours, storage GB) and disclosed.

Break-even unit volume = fixed cloud cost ÷ (revenue per unit − variable cloud cost per unit).

Compare unit cost against pricing: if unit cloud cost approaches price per unit, margin is eroding and pricing or architecture must change.

Report unit economics alongside absolute spend; a rising bill with falling unit cost can be healthy growth.