Realized Savings vs Identified Savings in Cloud FinOps A dashboard flags a $2.2 million optimization opportunity. The team celebrates. Then finance asks a simple question: how much of that actually reduced the bill?

That question exposes a gap many FinOps teams struggle to close. Flexera's 2026 data puts wasted IaaS and PaaS spend at 29%, which frames the opportunity but says nothing about what was achieved. Identified savings guide prioritization and forecasting. Realized savings prove value.

This article compares the two metrics: definitions, evidence, why the gap appears, how to measure it, and when to use each.

Key Takeaways

  • Identified savings are estimates; realized savings are validated outcomes after implementation.
  • Priorities come from identified savings; bill impact comes from realized savings.
  • A credible claim needs a baseline, a documented action, and billing evidence.
  • Growth, partial remediation, and commitment underuse widen the gap.
  • Mature teams track both metrics through one lifecycle.

Realized Savings vs Identified Savings: Quick Comparison

Dimension Identified savings Realized savings
Status Potential or forecast reduction, not yet validated Portion confirmed through actual cost, usage, or unit-economics data
Evidence Opportunity source, baseline, proposed action, assumptions, expected value Billing records, usage data, resource changes, reporting period
Timing Recorded when an opportunity is detected and valued Recorded after implementation and enough post-change data
Primary use Build and prioritize the backlog, forecast value, assign owners Report financial impact to finance and executives
Main risk Overstated when usage is assumed stable or adoption immediate Understated or hard to attribute amid growth, shared costs, commitment changes

Provider tools show why identified numbers need caution. AWS describes its estimated monthly savings as a "quick approximation" of future savings, normalized to 730 hours and dependent on later usage. Azure warns that some cross-series rightsizing actions can raise costs under existing reservations.

What Are Realized Savings and Identified Savings in Cloud FinOps?

What are realized savings?

Realized savings are the validated financial or unit-cost improvement after an optimization action, measured against a baseline agreed before or at the start of the initiative.

Evidence typically includes:

  • Billing exports (AWS CUR 2.0, Azure Cost Management exports, Google Billing export)
  • Usage reports and resource configuration history
  • Allocation data and change records

Realized savings differ from related concepts. Cost avoidance prevents future spend that never appeared in a baseline. Projected savings are forecasts. Efficiency gains that don't move the measured cost metric aren't savings at all.

Use cases of realized savings

Each action needs its own proof:

  • Deleting idle resources: Unattached EBS volumes stay chargeable until removed, so the line item should disappear from the bill.
  • Rightsizing compute: Compare post-change cost with the baseline, accounting for discount interactions that can produce $0 savings.
  • Overprovisioned block storage: EBS charges follow provisioned size and type, so a smaller provisioned footprint should show in billing.
  • Changing storage tiers: Look for the rate change on the affected volumes.
  • Commitment utilization: Check utilization, not just coverage. Azure amortized data exposes unused reservation cost.
  • Autoscaling: Compare actual versus estimated impact, since provider guidance alone isn't a quantified result.

What are identified savings?

Identified savings are the estimated value of an opportunity before the outcome is confirmed. They rest on assumptions about usage, rates, resource behavior, and implementation scope.

They help FinOps, engineering, and finance decide where to act. But an opportunity isn't realized savings until the action and measurement period are documented.

Use cases of identified savings

Teams rank opportunities across compute, storage, databases, networking, and commitments by:

  • Expected value
  • Effort and risk
  • Owner
  • Time to realization

Good estimates also include a confidence level and validation conditions, such as approvals, workload testing, tagging coverage, or minimum observation periods.

Realized Savings vs Identified Savings: What Is Better?

Neither. Identified savings tell you where to act. Realized savings prove the result. A mature program needs both.

Agree on definitions first

Build a shared glossary covering identified, forecasted, run-rate, realized, avoided, and efficiency savings. Finance and technical stakeholders should agree on treatment before optimization begins. The FinOps Foundation separates cost avoidance, savings, and cost reduction explicitly, and your glossary should too.

The savings lifecycle

  1. Detect the opportunity.
  2. Validate the baseline.
  3. Approve the change.
  4. Implement it safely.
  5. Compare post-change billing and usage.
  6. Adjust for external factors.
  7. Report the realized result with evidence.

Seven-stage cloud FinOps savings lifecycle from detection to reporting

Why identified savings fail to become realized

  • Partial implementation
  • Workload or resource growth
  • Changed usage assumptions
  • Commitment underutilization (for example, negative CUD savings on Google Cloud)
  • Configuration drift, since rightsizing is ongoing
  • Unallocated shared costs
  • Actions rejected over performance or reliability risk

What to report

Track these figures separately:

  • Gross opportunity
  • Implemented opportunity
  • Realized savings
  • Remaining run-rate potential
  • Variance to forecast

Useful metrics from FinOps Foundation guidance include:

  • Proportion of recommendations acted upon
  • Average time to resolution, which should trend downward
  • Forecast variance: (forecast − actual) / forecast
  • Effective Savings Rate, which applies only to RI and Savings Plans
  • Unit cost, such as cost per GB or vCPU
  • Allocation quality

Six cloud FinOps metrics for measuring savings realization and efficiency

No universal dollar-realization formula exists, so define yours and document it. Net savings should also subtract implementation costs such as tooling and engineering time.

Real-world Cloud FinOps Examples and Case Studies

Sky: discovery versus outcome

In Google's Sky case study, Sky found more than $1.5 million in immediate savings over six months after June 2019. In 2020 it reported more than $3 million on Google Cloud plus more than $800,000 on other providers. Teams used Cloud Billing, BigQuery, Looker, and Google recommenders, then followed through with engineering action.

Sky Google Cloud case study showing savings outcomes across reporting periods

One caution: discovery and outcome periods and scopes don't match, so no conversion ratio can be drawn. Its NowTV unit implemented more than 90% of recommendations in three weeks, a customer-specific result rather than a benchmark.

GE Vernova: a target is not an inventory

AWS's GE Vernova case reports more than $1 million cut in under a year against a $1 million target. Methods included lifecycle automation, idle-instance decommissioning, rightsizing, and database optimization. The total identified opportunity was not disclosed.

Block storage: identified versus realized

Lucidity's Lumen shows how an identified figure looks. Its published example lists:

  • $800K annual cost
  • $310K optimized cost
  • $490K (39%) potential annual savings

That is an identified number, with no billing evidence or normalization method attached, and it should be reported that way.

Lumen identifies four idle-disk types that often don't appear in native cloud dashboards and can represent up to 70% of unused block-storage spend:

  • Unattached disks
  • Reserved disks
  • Unmounted disks
  • Zero-I/O disks

Recommendations draw on historical IOPS, throughput, latency, and cost trends. Tier changes can be scheduled, run in one click, or rolled back.

On the realized side, Dometic reported a 52% reduction in cloud storage spend after adopting Lucidity AutoScaler, which resizes disks with zero downtime. Treat that as a customer-reported outcome, and verify your own against billing data.

Cloud storage savings examples comparing identified potential with realized reduction

Questions to ask your own numbers

Are your savings claims still opportunities, or can they be traced to implemented changes and validated billing outcomes? A quick baseline helps answer that. Lucidity's free Assessment scans an environment in about five minutes with no installation, giving block-storage teams a starting point for that review.

Conclusion

Identified savings are a planning signal, not proof that money has been saved. Realized savings need a defensible baseline, a completed action, a suitable measurement period, and evidence tied to actual costs or unit economics.

Use identified savings to prioritize the roadmap. Use realized savings to report outcomes. Connect both through a monitored lifecycle so optimization stays financially credible and operationally safe.

Frequently Asked Questions

How can I show proof of savings?

Link an agreed baseline and a documented action to post-change billing, usage, allocation, or unit-cost data. Adjust for workload growth and other external factors.

What is the difference between identified savings and realized savings?

Identified savings are estimated opportunities before validation. Realized savings are confirmed outcomes after implementation and measurement.

How do you calculate realized savings in FinOps?

Compare actual post-optimization cost with a documented, normalized baseline. Then adjust for usage changes, commitments, timing, and implementation costs. For Google CUDs, the formula is on-demand cost minus net cost.

Why do identified cloud savings fail to become realized savings?

Common causes are incomplete remediation, changing demand, drift, commitment or allocation issues, wrong assumptions, and actions delayed or rejected for reliability reasons.

Which FinOps metrics should track identified and realized savings?

Use realization (actioning) rate, time to realize, forecast variance, net savings, unit-cost change, and remaining optimization potential.