In July, AWS put a deceptively simple number onto Billing and Cost Management dashboards: cost efficiency. It can be tracked over time, filtered by account or Region, exported to scheduled reports, and placed beside budgets, Cost Explorer, and commitment coverage.
The score answers a useful question: how much of the AWS spend that Cost Optimization Hub considers optimizable remains represented by a current savings recommendation? AWS calculates it as one minus potential savings divided by total optimizable spend. If supported services account for $100,000 of qualifying monthly spend and the hub finds $10,000 of deduplicated opportunity, the score is 90 percent.
That clarity will make the metric popular with executives. It also makes misuse almost inevitable. A 95 percent score is not proof that cloud is creating value. A five-point improvement is not necessarily realized savings. A low-ranked account may be making a rational resilience tradeoff, while a high-ranked account may run an expensive architecture outside the metric’s field of view.
The score belongs on the dashboard. It should not become a grade for teams, a bonus target, or the sole definition of FinOps maturity.
Read the denominator before the result
Total optimizable spend is not total AWS spend. It includes spend in services for which Cost Optimization Hub supplies recommendations, including EC2, ECS, EKS, EBS, RDS, SageMaker, Redshift, Lambda, OpenSearch, MemoryDB, DynamoDB, ElastiCache, and related supported resources. AWS uses net amortized cost after credits and refunds and includes the entire service spend for supported services.
That boundary matters. Two business units can have identical scores while one has large data-transfer, support, marketplace, or service costs outside the denominator. A replatforming may move spend into or out of supported categories and change the score without changing operational discipline.
Publish the score with total optimizable spend, total AWS net amortized cost, and the coverage ratio between them. A 96 percent score covering 40 percent of spend communicates something very different from a 92 percent score covering 85 percent.
The metric also combines a rolling 30-day spend denominator with today’s savings opportunity. New resources, seasonal changes, and recommendation refreshes can move numerator and denominator on different rhythms. Annotate material launches, migrations, and business events on the trend.
Potential savings is a model, not cash
Cost Optimization Hub deduplicates overlapping recommendations and incorporates account-specific pricing and commitments. That is materially better than adding raw list-price suggestions. Stopping an idle instance, for example, reduces the commitment opportunity attributed to the same usage.
Yet the output remains an estimate normalized to a 730-hour month. Actual savings depend on future demand, implementation timing, resource replacement, commitment coverage, and whether the organization spends the released capacity elsewhere. An engineer may rightsize a database, only for growth to require an upsize two weeks later. The recommendation was implemented; the annualized headline was not realized.
Create three ledgers. “Identified” records the provider estimate at decision time. “Implemented” records the approved technical change and date. “Realized” compares a normalized baseline with actual net amortized cost after controlling for demand, price, and architecture changes. Never sum all three.
A higher score can conceal a worse decision
Any metric linked to performance invites gaming. Teams can defer provisioning until after a reporting date, delete useful headroom, buy commitments to suppress visible opportunity, or avoid supported services. They can also reject investments whose unit economics are excellent because absolute cost creates recommendations.
Pair efficiency with service-level and business metrics: availability, latency, deployment throughput, customer transactions, gross margin, or cost per inference. A rightsizing action that saves 8 percent but increases checkout abandonment destroys value. An account whose cost doubles while cost per active customer falls by 30 percent may be performing exceptionally.
Do not set one universal threshold. Production transaction systems, development sandboxes, batch analytics, and disaster-recovery environments hold headroom for different reasons. Define peer groups and exception policies. A deliberate exception with an owner, rationale, and expiry is evidence of governance, not failure.
Use the score to ask better questions
At organization level, the trend reveals whether opportunity is being created faster than teams resolve it. At account level, sort by potential savings as well as percentage; a 60 percent account with $400 of opportunity should not outrank a 91 percent account with $200,000. At Region level, investigate architecture and data placement before blaming ownership.
Decompose movement by action type: idle deletion, rightsizing, Graviton migration, storage change, and commitments. A score driven upward by a large Savings Plan purchase has different future risk from one improved through eliminated waste. Monitor coverage and utilization separately so rate optimization does not hide stranded commitment.
The new dashboard export and cross-account sharing make a monthly control practical. Archive the underlying opportunity set with each reported score. Otherwise, a reviewer can see that the number changed but not which recommendations entered, disappeared, or were superseded.
A board-ready scorecard
Present cost efficiency as one line in a compact value scorecard:
- Provider-modeled efficiency, opportunity dollars, denominator coverage, and 90-day trend.
- Identified, approved, implemented, and finance-validated savings, kept as distinct stages.
- Commitment coverage and utilization with expiry exposure.
- Unit cost and demand for the business service.
- Reliability guardrails and expired exceptions.
This format lets a leader see whether the estate is becoming leaner, whether work reached the invoice, and whether the technology delivered more value. No single number can answer all three.
A 30-day adoption plan
Enable Cost Optimization Hub and Compute Optimizer coverage across the organization, then place overall and account-level widgets on a FinOps dashboard. In week one, reconcile the denominator to the bill and document exclusions. In week two, assign the largest deduplicated opportunities by accountable team and capture accept, reject, or defer decisions.
In week three, connect completed recommendations to deployment records and begin measuring actual cost after change. In week four, publish the paired scorecard, segment accounts by workload type, and review anomalous movements rather than ranking teams.
Set a policy that the score cannot be used for individual performance evaluation without business and reliability context. That one sentence prevents the most predictable damage while preserving the metric’s value.
Visibility without false certainty
AWS has made a fragmented optimization estate easier to read. Daily refreshes, deduplicated opportunities, account and Region views, and scheduled exports are useful progress. The mistake would be asking the number to certify more than its model measures.
Cost efficiency is a map of currently visible opportunity. Finance must still verify the destination, engineering must still protect performance, and product leaders must still explain what the spend produced. Use the score to focus that conversation—not to end it.
Sources and reporting notes
- AWS announcement: Cost Efficiency widget for Billing and Cost Management dashboards, July 16, 2026.
- AWS documentation: Understanding your cost efficiency metric, including formula, denominator, and refresh basis.
- AWS documentation: Estimating monthly savings, consulted for normalization and realized-savings caveats.
- AWS documentation: Cost Optimization Hub, consulted for aggregation and deduplication behavior.
CostDefender turns provider recommendations into an accountable savings workflow—connecting opportunity, owner, implementation, invoice evidence, and the unit economics that show whether efficiency created value.