Chevron cut fifteen million dollars from Oracle support without dropping coverage on a single running system. How the reduction was landed at the renewal table, and the written controls that kept it from creeping back.
Chevron saved fifteen million dollars on Oracle support through a license utilization assessment, support line terminations, and contract cost controls applied across a multi year program. This page owns the closing chapter of the series: how a reduction is landed with Oracle at the table, and which written controls keep it from creeping back.
A material share of the support bill covered licenses that no longer ran anything. Years of platform consolidation onto Exadata had retired older Database deployments, but the support lines behind them kept renewing at Oracle's standard support rate, about 22 percent of net license fees, each year.
Oracle support renews by inertia. Every license ever bought generates a support line that renews annually unless someone acts, and the technical support policies are written to make acting expensive.
Fragmented CSIs mean fragmented renewal dates, fragmented account coverage, and no single view of the total. Oracle negotiates the whole relationship; a buyer negotiating line by line across scattered paper brings a knife to an artillery exchange.
Consolidating the paper was therefore not administrative housekeeping. It was the precondition for negotiating as one customer with one number.
The assessment matched every support line to a running workload and found the gap. Licenses with no deployment, options licensed but disabled, and capacity bought for projects that never shipped all surfaced in one consolidated view.
Support spend findings by category
| Category | Share of support spend | Action taken |
|---|---|---|
| Active production licenses | Majority | Retain and renegotiate |
| Licenses with no workload | Material minority | Terminate support lines |
| Options unused on licensed nodes | Smaller share | Terminate at license set boundary |
| Duplicate coverage across CSIs | Smaller share | Consolidate then terminate |
Because Oracle's matching service level policy holds every license in a license set to the same support level, and repricing rules can raise the unit cost of what remains after a partial reduction. The saving survives only when terminations follow set boundaries; the companion Costco case dissects those mechanics, and the LVMH case covers the evidence method feeding them.
You land it at renewal dates, with the model finished before the account team hears a word, and with every concession captured in writing before signature. Oracle rarely concedes support price; what a prepared buyer wins is structure, and structure is where the money was.
After the internal position is final and inside the planning horizon of the notice window, not before. Intentions floated early become retention campaigns: escalations, bundled offers, and delay, all aimed at reaching the renewal date with nothing decided.
Disclosure discipline is not hostility. It is the same professionalism Oracle applies in reverse, and account teams recognize a buyer who understands the game.
Four arrive so reliably they can be rehearsed. Each is legitimate commercial play, and each has an answer that keeps the program on course.
The standard counters and the answers that hold
| Oracle's counter | What it is doing | The answer that holds |
|---|---|---|
| Cloud credits or migration incentives | Converting a cost cut into new committed spend | Evaluate cloud on its own business case, never as a support offset |
| A ULA or bundled agreement proposal | Absorbing the cleaned estate into a bigger commitment priced off old spend | Price it against the reduced base and decline politely if it grows the total |
| Audit signaling | Testing whether the buyer's evidence is real | Hold the entitlement baseline ready and keep the tone commercial |
| Verbal assurances of future flexibility | Trading paper value for goodwill value | Thank them, then put every number in the order document |
Five items: confirmed pricing on surviving support lines, the uplift cap and its duration, the effective date of every change, the surviving discount structure on future purchases, and the list of terminated sets. Our support renewal contract checklist details the clause language worth holding out for.
Anything agreed in a call and absent from the paper does not exist. That sentence has cost buyers more than any policy Oracle publishes.
Nothing binds between renewals, so the calendar is the real clock of the negotiation. Each conversation was anchored to a specific renewal date, with the notice window counted backward from it as the deadline for internal decisions.
Buyers who negotiate without a date negotiate forever. Oracle's account teams are measured quarterly and respond to deadlines they can see; an open ended discussion suits only the side collecting the invoice meanwhile.
One page: the current support spend, the evidence summary behind the reduction, the sets leaving at the next renewal, and the written confirmations requested. No threats, no justifications beyond the evidence, no negotiable padding.
The brevity is the message. A single page that is obviously the surface of a deep file signals preparation better than the file itself ever could.
Three internal roles, used in sequence: procurement speaks, licensing supports, and the executive sponsor appears exactly once. Oracle fields its own ladder in response, and matching the rungs deliberately keeps the buyer in control of the tempo.
Rising seniority on Oracle's side of the table is progress, not pressure. Account representatives cannot approve structural concessions; the appearance of their management usually means the position paper has been read upward and taken seriously.
The error to avoid is celebrating the attention by softening the ask. The position that earned the meeting is the position that should finish it.
A live alternative is negotiating oxygen, whether or not it is exercised. A current third party support quote for eligible tranches of the estate converts "we could reduce" from an aspiration into a priced plan, and Oracle's side can tell the difference immediately.
The alternative must be real enough to execute. A quote requested purely as theater reads as theater, and it undermines the credibility everything else was built on.
Three moves carried the program: terminate support on unused license sets, consolidate contracts so terminations priced cleanly, and apply repricing caps in writing at each renewal. None of them required dropping coverage on a single running workload.
Because consolidation changes what a termination does to the surviving paper. On fragmented CSIs, each reduction is priced against its own small contract history; consolidated, the surviving estate negotiates as one position with one renewal conversation.
Consolidation also surfaces duplicate coverage that fragmentation had hidden, which in this program was itself a savings category. The overlap only became visible when the paper sat in one place.
Paper, not goodwill. The lifetime support policy and the repricing rules are Oracle's levers; written caps and clean license set terminations are the buyer's. Every saving in the program was structured to survive both.
The uplift cap and the repricing confirmation, in that order. The uplift cap bounds the annual increase on the surviving base for a defined term; the repricing confirmation states that the reduction did not change surviving rates.
Wider renewal protections belong in the master negotiation, covered in our Oracle contract renewal strategy guide. The support caps are the minimum a reduction program should never sign without.
Three policy mechanics decide what a cost control program can and cannot promise: the Sustaining Support transition, the reinstatement formula, and the timing rules around the annual uplift. Each was priced into the program before anything was promised to finance.
Because the fee does not fall when a product ages out of Premier; the rights do. Sustaining Support keeps invoicing at the full rate while new patches, fresh certifications, and new version access stop arriving.
A governance program treats that transition as a decision point, not a default. When a product nears the end of its Premier window, the choices are upgrade, renegotiate, move eligible systems to another support model, or knowingly pay full price for a thinner service; the current rate card and uplift picture sits in our Oracle support costs guide.
Reinstatement starts from the last annual support fee you paid and adds back support at 150 percent for the lapsed period under Oracle's policies. It is designed to cost more than never having left.
That formula is not a reason to avoid terminations; it is the number every candidate set had to beat. The mechanics, and when reinstatement is ever worth paying, are covered in our guide to dropping Oracle support and reinstatement.
At renewal, before signature, and at no other moment. Once the renewal paper is signed the uplift is contracted for the term, and no amount of relationship goodwill reopens it midstream.
That timing rule is why the calendar discipline earlier on this page matters as much as the evidence. Every renewal date is a scheduled chance to cap the uplift; a missed date is a full year at the uncapped rate, invoiced while you wait for the next opening.
The program delivered fifteen million dollars in support savings while every running workload kept full support. The savings came from scope and structure, not from service level downgrades.
With two standing controls: a purchase gate and an annual review. Every proposed Oracle purchase passes through one owner who prices its permanent support tail before approval, and every renewal is preceded by a utilization pass against the set map.
The controls also manage audit posture, since estate changes are among the events that draw attention; our guide to what invites an Oracle audit covers the signals worth managing deliberately.
Three questions before any order: does an existing entitlement already cover the need, what does the support line cost over five years, and which license set will the purchase join or create. Thirty minutes of review per purchase protects the entire program's arithmetic.
The five year framing matters most. A license priced at its purchase cost looks cheap; priced with its permanent support tail, it competes honestly against alternatives.
That the support line deserves the same scrutiny as the license line. Most enterprises negotiate licenses hard and renew support blind, and the blind renewal is where Oracle's margin lives. The phased program that industrializes this discipline is the companion American Airlines case.
Recognizable, because it is common: scattered cancellations without a set map, savings announced before repricing was modeled, verbal assurances accepted at renewal, and no purchase gate afterward. The gross saving would have looked similar in the announcement deck.
The difference appears in year two, when repricing and creep quietly return a third or more of the money. Structure is not overhead on the saving; structure is the saving.
The standard advice says walk in with the termination threat and let leverage do the work, or alternatively that Oracle support is untouchable so negotiating is pointless. We disagree with both poles. Across the 30 to 40 engagements Fredrik Filipsson advised in 2024 and 2025, threats without finished evidence collapsed under the first audit signal, while fatalism left 10 to 25 percent of support spend covering nothing year after year. What actually moved Oracle was a completed model, clean set boundaries, and a buyer visibly prepared to execute without a deal. Leverage is not the threat; leverage is the evidence that the threat is already priced.
Three numbers set the boundary of what disciplined negotiation protects.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
The checklist below sets up both halves of this case: landing the reduction, then defending it.
White Paper · Oracle
When third party support is the right call for Oracle Database, Apps, and Middleware. Read it free.
By terminating support on license sets with no deployed workload, consolidating contracts so terminations priced cleanly, and capping repricing on the surviving base. Coverage on running workloads never changed.
Rarely on the rate itself. What prepared buyers win is structure: which sets terminate, what the surviving base pays, and how fast it can grow. Treat the rate as fixed and negotiate everything around it.
A written limit on the annual support increase for a defined term, agreed at renewal. It is a routine ask for buyers negotiating with evidence in hand, and it is worth more than most one time concessions because it compounds.
Only after the internal model is finished, inside the planning window for notice. Early disclosure hands the account team a full cycle to run retention plays against an unfinished position.
It changes the negotiation, not the coverage. A cleaner, smaller support base entering each renewal strengthens the buyer position, and account teams engage more seriously with customers whose numbers are defended.
Two: a purchase gate that prices the five year support tail of every new order, and an annual utilization pass before each renewal. Together they keep the support base tracking the running estate instead of the purchase history.
It helps materially, provided it is genuine. A priced, executable alternative for eligible systems anchors the conversation in consequences rather than requests; an obviously tactical quote does the opposite.
The support optimization sequence, the license set rules, and the repricing traps.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement leaders running the next renewal cycle.
Oracle support renews by inertia. The estates that audit the support line annually are the ones that never pay for nothing.
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.
