Savings Plans optimization: $6.4M across 12 accounts.

The customer ran 12 AWS accounts behind a single AWS Organizations structure with no shared payer-level discount strategy. Each engineering team had purchased Reserved Instances independently over a three-year period, creating a portfolio of 287 individual RI line items with mismatched terms, expiration dates, and instance families. Roughly 40 percent of the RIs were under-utilized; another 18 percent had expired and quietly rolled to on-demand pricing.

The original purchase pattern made sense locally — each team had bought what it needed when it needed it — but the result at the organization level was a fragmented portfolio that left 49 percent of steady-state compute uncovered while simultaneously paying for unused RI capacity. The CFO had asked the FinOps team for a one-page explanation; the FinOps team had concluded the structure was beyond reasonable internal capacity to optimize.

What the customer needed

• A clean view of the full commitment portfolio across all 12 accounts
• A migration path from legacy Reserved Instances into the optimal Savings Plan structure
• Coverage targets that matched the actual steady-state vs. burst consumption profile
• A governance framework so the new structure would not fragment again over time

Savings Plans optimization is mostly a data problem. The negotiation lever is small compared to a fresh EDP; the savings come from getting the commitment shape, term mix, and coverage right against actual consumption.

Phase 1 — Consolidate and model (weeks 1-2)

We pulled 18 months of hourly consumption data across all 12 accounts and built a unified consumption model at the payer level. The model produced a true steady-state load, a burst envelope, and a probability distribution for new workloads coming online based on the customer's product roadmap.

Against that model, we ran 14 candidate commitment structures, varying the Compute-vs-EC2 plan mix, 1-year vs. 3-year term split, and partial-vs-no-upfront payment options. The optimal structure produced 94 percent coverage on steady-state load at the lowest effective rate.

Phase 2 — Migration path (weeks 3-4)

The legacy RIs had different expiration dates spread across 30 months. We built a phased migration where each new Savings Plan purchase was timed to coincide with RI expirations — never overlapping coverage, never leaving gaps. Twenty-three RIs that still had material time remaining were modeled for sale on the RI Marketplace; eleven were sold, recovering $480K in residual value.

Phase 3 — Governance and execution (weeks 5-6)

The structural change was paired with a governance update: all future commitment purchases route through a central FinOps review against the unified consumption model, instead of being made by individual engineering teams. The customer's existing FinOps tool was reconfigured to enforce the new policy.

The restructured Savings Plan portfolio produced $6.4M in two-year savings against the trajectory the customer was on. The savings split across three buckets and exposed a fourth opportunity for a future cycle.

Where the savings came from

• Coverage uplift — $3.6M from raising effective coverage on steady-state load from 51% to 94%, eliminating the on-demand premium on workloads that were running 24/7
• RI rationalization — $1.9M from retiring under-utilized RIs and replacing them with Savings Plans sized to actual consumption
• Term and payment optimization — $0.9M from shifting roughly 60% of the portfolio to 3-year partial-upfront, the optimal trade-off given the customer's cash position and consumption certainty
• Future opportunity identified — the customer is approaching a $5M+ annual spend threshold where EDP-level Private Pricing Addenda become viable. We have flagged this for the next engagement cycle.

What the customer did with the savings

The CFO redirected most of the freed compute budget into the customer's first material AI workload commitment, which had been deferred for two quarters due to cost concerns. The remaining savings absorbed compute growth from a new enterprise tier of the product without requiring a budget revision.

The governance structure is the long-lived change. The customer has not added a single individual Reserved Instance since the engagement closed; all commitment activity now routes through the central FinOps process against the unified consumption model.