Oracle does not send one bill, it sends five
Database on Processor or Named User Plus, middleware on Processor, applications on user or employee metrics, Java on total employees, cloud on credits, each under contract paper that is not always the same document. A buyer who fixes the database line and leaves the other four running has moved a quarter of the problem: this is the orientation layer across the whole estate.
Prepared by Redress Compliance · August 7, 2026 · Oracle advisory. Based on 35 to 45 Oracle estate reviews run 2024 to 2025.
Executive summary
The surplus is real, and it is never in one place.
The median estate carried 25 to 35 percent more license than its signed contracts required: unmapped option usage, most often Partitioning and Diagnostics Pack, appeared in seven of ten estates and added 10 to 20 percent to the compliance gap each time.
Named User Plus counts sat at the 25 per Processor minimum where actual users numbered under ten, inflating cost two to five times on small databases; and middleware was absent from the internal license position more often than present, while never absent from the audit scope.
Policy documents are not contract, and pricing to them buys license nobody owes.
The partitioning policy and the core factor table are referenced as policy, not signed as contract, and buyers sizing to Oracle policy documents rather than their signed agreements over provisioned by 15 to 30 percent, in roughly six of ten estates we reviewed.
The audit clause itself is one paragraph naming no tool: 45 days notice and access to information, not a method, a deadline, or your hypervisor inventory.
The options are the bill, not the engine.
Six common Database options, Partitioning, Diagnostics, Tuning, Advanced Compression, Advanced Security, and RAC, list at $73,500 per Processor combined against the $47,500 Enterprise Edition engine underneath: on a 16 Processor server that is $1,176,000 of options standing on $760,000 of database.
A 1.55 to 1 ratio, and every option lists on Named User Plus at one fiftieth of its Processor price, so both columns move together.
Fixing the engine metric and leaving the option lines alone never moves the number.
Support falls only when the base falls. Support is charged on the licensed base, so nothing aimed at the support line alone moves it: take out the unused options and the surplus capacity and support falls with them.
Java runs the same logic in reverse, priced on every employee rather than Java users, where the first count buyers put forward ran 18 to 28 percent above what they could defend, and the five contract vehicles carry different audit language, termination rights.
And subsidiary definitions, which is why the single most useful hour in an Oracle program is writing all five lines on one page with their metric and their paper next to them.
The five cost lines, on one page
| Line | Metric | What moves it | Contract vehicle |
|---|---|---|---|
| Database | Processor or Named User Plus | Options, packs, and the platform underneath | Master agreement and ordering document |
| Middleware | Processor or Named User Plus | Edition tier and restricted use boundaries | Master agreement and ordering document |
| Applications | Application user, employee, or module | User breadth and module count | Applications price list, separate paper |
| Java SE | Every employee, tiered | Total headcount including contractors | Subscription order, often outside the OMA |
| Cloud | Universal Credits or BYOL | Commitment size against actual draw | Cloud services agreement and order |
The last column is the one people skip. Different vehicles carry different audit language, different termination rights, and different definitions of who your subsidiaries are, so read them as five contracts, because that is what they are.
Nothing learned on one line transfers cleanly to another: the Processor arithmetic multiplies physical cores by the core factor table, 0.5 for Intel so a 32 core server needs 16 licenses rounding up, while cloud counts vCPUs with no factor at all.
And employee metrics move only by contract definition, never by anything you do technically.
The options catalog, where the bill actually grows
| Option or pack | List per Processor | 16 Processors at list | Where the use is recorded |
|---|---|---|---|
| Partitioning | $11,500 | $184,000 | DBA_PART_TABLES and the feature usage view |
| Diagnostics Pack | $7,500 | $120,000 | AWR snapshots and the management pack parameter |
| Tuning Pack | $5,000 | $80,000 | SQL Tuning Advisor tasks in DBA_ADVISOR_TASKS |
| Advanced Compression | $11,500 | $184,000 | COMPRESS_FOR reading ADVANCED in DBA_TABLES |
| Advanced Security | $15,000 | $240,000 | Encrypted tablespaces and columns views |
| Real Application Clusters | $23,000 | $368,000 | GV$INSTANCE returning more than one row |
| All six combined | $73,500 | $1,176,000 | Against a $760,000 engine |
The Oracle CIO complete playbook
All five lines worked end to end: the metric decisions, the option audit, the contract paper map, and the negotiation calendar.
Get the white paper →The metrics, and the minimums that catch buyers
- Processor fits large or unknown user populations and anything internet facing: physical cores times the core factor, rounded up, worked in the core factor analysis.
- Named User Plus counts every human and device authorized to access, not active users, with multiplexing through an application server reducing nothing.
- The minimums: Enterprise Edition requires 25 Named User Plus per Processor, so a four Processor server needs 100 named users even if ten log in; Standard Edition 2 floors at 10 per server.
- The pre signature move: model both metrics and put both in the ordering document as priced alternatives, because a metric switch alone can never save more than half.
- The employee metrics move only by contract definition, which is why the Java count, first drafts running 18 to 28 percent above defensible, is a negotiation and not a census.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across estate reviews, 2024 to 2025
Across roughly 35 to 45 Oracle estates Fredrik Filipsson reviewed between 2024 and 2025, the median estate carried 25 to 35 percent more license than its signed contracts required, and the surplus was never in one place:
On small databases counted at the 25 per Processor floor where actual users numbered under ten.
Over provisioned by buyers sizing to the partitioning policy and core factor table instead of contracts.
The middleware line deserves its own sentence: absent from the internal license position in more estates than it was present, never absent from the audit scope, with WebLogic shipping restricted use inside applications and converting to full use the moment anything else touches it.
The deep dives sit one layer down, the database licensing reference for the engine and options arithmetic, the ULA lifecycle where unlimited terms reshape all five lines at once, the EBS applications guide, and the Support Rewards arithmetic, the one mechanism that moves the support line down.
Your first five moves
- Write the five lines on one page, each with its metric and its contract vehicle, the single most useful hour in an Oracle program.
- Audit option usage against the recording objects, because seven of ten estates carried unmapped Partitioning or Diagnostics use.
- Size to your signed contracts, never to policy documents, the error that bought unowed license in six of ten estates.
- Model both metrics with the minimums before signing, where the small database floor inflates two to five times.
- Cut support by cutting the base: unused options and surplus capacity out, support falls with them. The Oracle practice runs the estate with you.
Frequently asked questions
How does Oracle licensing work?
As five separate cost lines on five metrics under different contract paper: database on Processor or Named User Plus, middleware on Processor, applications on user or employee metrics, Java on total employees, and cloud on Universal Credits.
Each vehicle carries its own audit language, termination rights, and subsidiary definitions, and nothing learned on one line transfers cleanly to another.
What is the difference between Processor and Named User Plus licensing?
Processor multiplies physical cores by the core factor table, 0.5 for Intel, rounded up, and fits large or internet facing populations.
Named User Plus counts every authorized human and device, not active users, with minimums of 25 per Processor on Enterprise Edition and 10 per server on Standard Edition 2.
Model both before signing: the floors inflated small database costs two to five times.
How much do Oracle Database options cost?
More than the engine: Partitioning at $11,500 per Processor, Diagnostics at $7,500, Tuning at $5,000, Advanced Compression at $11,500, Advanced Security at $15,000, and RAC at $23,000, totaling $73,500 against the $47,500 Enterprise Edition engine, a 1.55 to 1 ratio.
Each records its use in specific database objects, which is what an option audit reads.
Are Oracle policy documents legally binding?
No: the partitioning policy and the processor core factor table are referenced as policy, not signed as contract, and buyers sizing to them rather than their signed agreements over provisioned by 15 to 30 percent, buying license nobody owed in roughly six of ten estates we reviewed.
Your ordering document and master agreement are the texts that govern; price against those.
How do you reduce Oracle support costs?
By reducing the licensed base, because support is charged on it and nothing aimed at the support line alone moves it: terminate the unused options and surplus capacity and support falls with them.
The one mechanism that moves the line down without shrinking the base is Support Rewards, which converts eligible OCI consumption into credit against technology support at 25 to 33 cents per dollar.
What does the Oracle audit clause actually allow?
One paragraph: 45 days notice and access to information. It names no tool, sets no method or deadline, and grants no right to your hypervisor inventory, which matters because audit practice routinely requests far more than the clause requires.
Middleware deserves particular attention, absent from most internal license positions and never absent from the audit scope.