Oracle
American Airlines: $12M off Oracle support over three years.
Unused license terminations, a roadmap based support split, and sequenced repricing math removed $12M of Oracle support cost. Here is how the program ran.
Unused license terminations, a roadmap based support split, and sequenced repricing math removed $12M of Oracle support cost. Here is how the program ran.
American Airlines cut Oracle support spend by $12 million over three years by terminating unused licenses, splitting the estate between Oracle and third party support, and pricing the repurchase risk before Oracle could. This page walks the program itself: how a three year effort is phased, gated, and governed so the number actually lands.
American Airlines carried a multimillion dollar annual Oracle support bill across database, middleware, and applications, with a meaningful share attached to licenses no longer deployed. Support renewals arrived as one roll forward number, growing at the standard uplift, with no internal map of which lines funded which running systems.
The airline industry context made the bill harder to defend. Cost programs after 2020 put every recurring line under review, and Oracle support was among the largest untouched items.
Cost pressure had already created executive appetite, and a cluster of support renewals inside the following eighteen months created the mechanical opportunity. Support changes only bind at renewal, so a program launched into a quiet stretch of the calendar has nothing to act on.
The practical rule: count the renewal dates in the next six quarters before proposing the program. That count is the program's real capacity.
Because the pricing rules punish partial moves. Under Oracle's technical support policies, matching service levels binds the license set, not the CSI, and a single CSI can hold several sets. Reduce part of a set and Oracle may reprice the support you keep, which is why casual line item cuts often save nothing.
Because the savings arrive in layers over three years, and each layer depends on the one before it. A one time exercise banks the easy terminations and stops; a governed program keeps compounding, which is where the second $8 million came from.
The setup ran in three phases over roughly nine months: inventory and entitlement mapping, termination sequencing against repricing exposure, and a support model split between Oracle and third party providers. The three year window that produced the $12 million then ran on top of that foundation, governed by an annual review cycle.
The three phase structure and what each delivered
| Phase | Core activity | Gate before proceeding | Contribution to the $12M |
|---|---|---|---|
| 1. Inventory | Map deployments to support lines and entitlements | Baseline agreed with finance | Found the unused 20 percent |
| 2. Termination | Sequence terminations around license set rules | Repurchase risk priced per line | Roughly 35 percent of savings |
| 3. Support split | Third party support on stable systems | Roadmap classification signed per system | Roughly 50 percent of savings |
| Ongoing | Annual roadmap review per system | Renewal calendar owned by one team | Stops the creep returning |
The baseline was the existing support spend projected forward with contracted uplift, signed off by finance before phase one closed. Every later claim of saving was measured against that trajectory, which ended the usual argument about whether avoided uplift counts.
Locking it early had a second effect: it converted the program from an IT initiative into a finance tracked commitment. Programs with a signed baseline get resourced; initiatives without one get deferred.
Each gate asked one question: is the downside of the next phase priced? Terminations did not start until repurchase exposure was modeled per line, and the support split did not start until every affected system carried a signed roadmap classification.
Gates slow nothing when the work is real. What they prevent is enthusiasm outrunning evidence, which in Oracle cost work is how savings turn into settlements.
Sustained effort from licensing, infrastructure, and procurement, plus external benchmark data where internal pricing knowledge ran out. The investment is real, and it is small against a $12 million return.
The honest warning: the setup produces no visible savings for months. Executive patience through that quiet period is part of the budget, and the locked baseline is what buys it.
Lines were terminated in an order that kept license sets whole, because breaking a set exposes the survivors to recalculation against the Oracle price list. The order came from modeling each candidate against the set map, never from the size of the line.
The full mechanics of set surgery, including why one CSI can contain several sets, are the spine of the companion Costco case. The evidence method that feeds it is covered in the LVMH case.
Stable systems with no planned Oracle upgrades moved to third party support at roughly half the Oracle rate. Systems on active roadmaps, where upgrades and patches from Oracle Premier Support carried real value, stayed on Oracle support deliberately. The evaluation criteria for that split are detailed in our third party support provider guide.
Middleware got the hardest look, because stable integration layers rarely consume new Oracle features yet often pay full Premier rates. Where a platform was frozen pending replacement, it became a support split candidate; the economics of going further and migrating off entirely are set out in the middleware migration business case.
The program removed $12 million of support cost over three years against the locked baseline, with the first full year delivering roughly a third of that. No production incident was attributed to the support model change across the measurement window.
Against the baseline trajectory, in three components: terminated fees that stop recurring, rate reductions from the support split, and avoided uplift on both. Each component was tracked separately so nothing was double counted.
The distinction matters at year three, when a new CFO asks where the $12 million lives. A program that can decompose its number survives that meeting; a single blended figure does not.
Front loaded by design. Terminations produce immediate, permanent reductions, so they went first; the support split then stepped the rate down in tranches as each system's renewal date arrived.
Sequencing tranches by renewal date is slower than switching everything at once, but it avoids paying twice on any line and keeps each move reversible until its own deadline.
Active roadmap systems stayed on Oracle Premier Support at full rate, without argument. Chasing every last dollar would have put upgrade paths and patch access at risk on platforms that genuinely needed them.
That restraint was strategic. A program that never overreaches is a program Oracle cannot point to when something breaks, and it kept the operational teams on side for three full years.
Oracle account pressure concentrated on audit signaling and cloud migration offers rather than support price relief, the standard response pattern under Oracle support policies. Holding documented entitlement records and a clean deployment map kept the audit conversation short.
The negotiation craft of landing a reduction, and the written protections worth demanding, are the spine of the companion Chevron case.
A small standing group owned the program across all three years: one licensing owner, one infrastructure owner, one procurement owner, and an executive sponsor with authority over the renewal calendar. The group met quarterly, and its agenda never changed.
One page per quarter: run rate against the baseline trajectory, the two quarter renewal outlook, open risks with owners, and any Oracle commercial contact. Nothing else.
Brevity was deliberate discipline. A program that reports simply gets read, and a program that gets read keeps its sponsor when budgets are challenged.
Every saving in the program landed on a renewal date, and every risk had a notice window attached. The team that owns the calendar owns the program; estates where renewal dates surprise people do not run three year programs, they run annual fire drills.
By treating the reduced base as a budget, not an achievement. The annual roadmap review reclassified every system each year, new purchase requests passed through the same governance group, and the baseline was rolled forward so drift became visible the quarter it started.
Programs decay when their founders move on. Writing the cadence into the operating rhythm, rather than into one person's calendar, is what made the third year look like the first.
Sponsor turnover, mid program temptations, and unguarded purchasing are the three killers we see. None of them is technical, which is why the defenses are all governance.
Failure modes and the defense that holds
| Failure mode | How it shows up | Defense |
|---|---|---|
| Sponsor turnover | New leadership inherits a program it did not sign | Finance owned baseline that survives personnel changes |
| Mid program ULA offer | Oracle proposes an unlimited agreement that would absorb the cleaned estate | Evaluate against the reduced base, never the old spend |
| Unguarded purchasing | A project buys licenses that quietly add recurring support | Route every Oracle purchase through the governance group |
| Audit panic | An audit letter freezes the program mid sequence | Entitlement documentation maintained from phase one |
| Baseline drift | Savings claims lose their reference point by year three | Roll the baseline forward annually, in writing |
An unlimited agreement pitched into year two is priced against your old spend, not your cleaned base, and it converts a shrinking cost line back into a growing one. Evaluate any such offer against the program's forward trajectory, and let the governance group, not the account team, frame the comparison.
The transferable lesson is sequencing. Most enterprises can find the same three savings layers, but the order of operations decides whether the savings survive Oracle's repricing and audit responses. The five rules below are the portable core of the program.
Yes, with smaller absolute numbers. The shelfware share and the roadmap mismatch exist at almost every estate over a few million dollars of annual Oracle support; the percentages in our engagement file hold from roughly $2M of support spend upward. Current rate benchmarks live in our Oracle support costs guide.
Three of them travel well: the unused share worth hunting for sits between 15 and 25 percent of the support line, year one should carry roughly a third of a three year target, and stable systems price at roughly half rate under third party support.
Treat them as planning assumptions to be verified against your own estate, not as promises. The verification is precisely what phase one exists to do.
The standard advice says bundle everything into one grand renegotiation: bring terminations, support splits, and renewals to the table at once and trade the package. We disagree. Across the 30 to 40 programs Fredrik Filipsson worked or benchmarked in 2024 and 2025, the single event approach consistently traded permanent structural savings for one time concessions, usually wrapped around a cloud commitment the buyer did not need. A phased program keeps leverage alive because Oracle faces a live decision every year, not one signature followed by silence. The buyer side move is to sequence, gate, and govern, and to let each year's results fund the next year's ambition.
Three figures from our advisory files size the opportunity a phased program is chasing.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Six moves start the program; the first quarter's job is the map and the baseline, nothing else.
White Paper · Oracle
Control Oracle Spend: The 5 Year CIO Playbook
The buyer side playbook for the CIO who controls Oracle spend across audit, ULA, Database, Java, OCI, and renewal in one five year plan. Read it free.
The program removed $12 million of Oracle support cost over three years, measured against a baseline that included contracted uplift. Terminations, a roadmap based support split, and third party support on stable systems each contributed a tracked share.
The foundation took roughly nine months across inventory, termination sequencing, and the support split; the full financial result accrued over three years. Faster is possible on smaller estates, but the gates should not be skipped.
Yes, when the license set math is done first. Matching service levels binds the set rather than the CSI, so each termination must be modeled against the set map before notice is served.
For systems with no planned Oracle upgrades, where stability matters more than new versions. Active roadmap systems usually justify staying on Oracle support, which is why the split is decided per system, per year.
It raises the probability of audit signaling, which is why the entitlement map and deployment records were built before any notice went out. A documented position keeps that conversation short and commercial.
Yes. The phasing, gates, and governance scale down cleanly; the percentages in our engagement file hold from roughly $2 million of annual Oracle support upward, with only the absolute savings changing.
Current support spend projected forward with contracted uplift, agreed with finance before the first termination. Without the uplift component the program undercounts its own result; without finance sign off the number will be relitigated every year.
A four role standing group meeting quarterly: licensing, infrastructure, procurement, and an executive sponsor who owns the renewal calendar. Its two standing powers are approving every new Oracle purchase and reclassifying every system's roadmap annually.
The termination sequence, repricing math, and support split model from 30 plus Oracle support engagements.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement leaders running the next renewal cycle.
Support optimization is a sequencing problem. The savings are obvious; surviving Oracle's repricing response is the craft.
500+ enterprise clients. 11 vendor practices. Industry recognized. One conversation can change what you pay for the next three years.
One buyer side briefing a week. Pricing moves, audit signals, and the levers that work. No vendor spin.
