The per employee metric is the only Oracle basis where a decision taken by corporate development, with nobody from IT in the room, changes the licensing bill on the day the deal closes
Once the counting basis is headcount, operational discipline stops paying. Removing the software from ninety percent of the estate changes nothing until it is gone from all of it.
Prepared by Redress Compliance · August 18, 2026 · Oracle advisory. 40 to 60 Oracle negotiations run or benchmarked, 2024 to 2025.
Executive summary
The per employee metric decouples price from usage entirely. That makes revenue predictable for Oracle and makes cost reduction almost impossible for the buyer, because the lever that used to work no longer connects to the bill.
It is the only metric where the risk owner sits outside IT. Processor risk sits with infrastructure and Named User Plus risk sits with application owners. Headcount risk sits with corporate development.
Partial exit stops working under it. Removing the software from most of the estate changes nothing at all until it is removed from every part of it, which is a different project with a different budget.
Verification becomes trivial for the vendor. Headcount is public or easily established, so a compliance conversation needs no estate data and no cooperation from you.
Which Oracle metric decides who carries the risk?
Four metrics carry most Oracle spend, and each one hands the risk to a different part of the business. The decision a CIO owns is not which price to accept, it is which risk to hold.
| Metric | Your bill grows when | Risk sits with | The control point |
|---|---|---|---|
| Processor | Hardware refreshes or clusters widen | Infrastructure | Architecture change approval |
| Named User Plus | Authorized populations grow or minimums bite | Application owners | Access governance and the joiner leaver process |
| Per employee | The organization grows or acquires | The whole company | Corporate development, not IT |
| Consumption credits | Engineering consumes, or the commitment expires unused | Engineering and finance jointly | A funded consumption forecast |
Which metric is right for a given workload is a separate question, worked in our brief on Oracle pricing metrics. This one is about who carries the risk once the basis changes.
The row that should stop a board
Per employee is the only metric where a decision taken with no involvement from IT changes the licensing bill on the day the deal closes. An acquisition is a licensing event under it, and nobody in the deal room knows that.
Under Processor licensing the count follows the estate, and under Named User Plus it follows the access list. Under the employee metric it follows the payroll, which is the one number IT has no influence over and no early warning about.
Why is the per employee metric spreading beyond Java?
Because it removes the counting problem for the vendor and removes the savings lever from the buyer at the same time. The Java SE Universal Subscription is the template the rest follows.
Oracle describes the model as per employee pricing that removes the need to account for installations across desktops, servers and clouds. That description is accurate, and it is exactly why the model needs careful modeling before it is accepted.
- Verification becomes trivial. Headcount is public or easily established, so no estate data is needed to open a compliance conversation.
- Growth is built in. Hiring and acquisition raise the bill with no change in technical usage at all.
- Partial exit stops working. Removing the software from most of the estate changes nothing until it is gone from all of it.
- The definition does the work. Contractors, agents and outsourced staff are usually inside the count, so the number is larger than the payroll number people picture.
Our Java licensing pillar works the employee count in detail, including how the definition is drawn and defended.
The Oracle CIO playbook
Metric counter models, the unbundling sequence, the support base defense, and the five year view that a spend forecast alone will always get wrong.
Get the brief →What 40 to 60 Oracle negotiations showed
Across roughly 40 to 60 Oracle negotiations Fredrik Filipsson ran or benchmarked in 2024 and 2025, the quiet driver of cost growth was the metric rather than the price. The rate moved a little. The counting basis moved a lot.
Metric change proposals arrived inside renewals, inside cloud migrations, and inside bundle restructures. They rarely arrived on their own, and they were almost never the headline of the document that carried them.
Oracle reported fiscal 2026 revenue of 67.4 billion dollars with cloud at 34.0 billion, up 39 percent, against software revenue of 24.5 billion, in its fourth quarter results announcement. Software revenue was 24.7 billion the prior year.
That matters to a buyer because the line most CIOs think of as their Oracle relationship is flat while the company grows. Growth has to come from somewhere, and a counting basis that rises with your headcount is one of the places it comes from.
Inside the agreements reviewed, 18 to 32 percent of bundled modules were inactive. That is the scope conversation, and it has to happen before the price conversation rather than after it.
- Your quote benchmarked against 500,000+ real closed deals, adjusted for size, region, and industry
- The proposed metric priced against the current basis across the full term, including growth
- Uplift and definition language flagged with the exact quote, the page, and the replacement text
How should a CIO respond to a metric change proposal?
Insist on a counter model before the commercial conversation opens, and make it a standing rule rather than a decision taken case by case. The rule is what protects you when the proposal lands in a busy quarter.
- Price the same coverage on the current metric across the full proposed term, with the growth assumptions written down.
- Price the removal path for the product, fully loaded, including engineering effort and the risk of stopping halfway.
- Test the sensitivity to a fifteen percent headcount change in each direction, because that is the range corporate plans actually move within.
- Name who owns the risk once the metric changes, and confirm that person knows they now own it.
Where the core factor still decides the number
Under Processor licensing the count is physical cores multiplied by a factor from the Oracle Processor Core Factor Table, which ordering documents reference rather than restate. On most x86 platforms the factor is 0.5, so a hardware standard set by an infrastructure team can move a licensing bill nobody consulted them about.
Our explanation of how the core factor works gives the arithmetic, and the technology price list gives the rates it applies to.
How should a CIO read an Oracle bundle?
Read it as a proposal about optionality rather than a proposal about price. A bundle almost always lowers the number you see this year and reduces the number of decisions you can take independently in the years after.
| Structure | What it gives | The option it removes |
|---|---|---|
| Unlimited license agreement | Deployment freedom during the term | The ability to shrink the support base at exit |
| Cloud commitment attached to a license renewal | Discount on the license line | Independent judgement on the cloud platform decision |
| Options and packs bundled with the database | A higher discount percentage | A clean support base and a low audit surface |
| Suite or applications agreement | One number and one renewal | Module by module repricing and partial exit |
Unbundle without losing the deal
Ask for component level pricing in writing, early, and as a normal part of your approval process. Framed as a governance requirement it is rarely refused. Framed as distrust it often is.
What the agreements measured, 2024 to 2025
Two cuts of the engagement file frame where the money moves.
Inside the agreements reviewed, which is the scope conversation that has to happen before the price conversation.
Across 2024 and 2025, where metric migration rather than rate increase was the quiet driver of cost growth.
Neither number is about discount. Both are about what is being counted and what is being carried, which is where a three year Oracle position is actually decided.
Watch the briefing · 4:47Employee and Revenue: The Metrics That Grow Without YouHow the employee metric counts, what sits inside the definition, and what it does to a partial exit.
Your first five moves
- Name the metric on every Oracle line you hold, and write down who carries the risk under each one.
- Tell corporate development that headcount is now a licensing input, since under the employee metric an acquisition raises the bill on the day it closes.
- Make a counter model a standing rule, so a metric change proposal is always priced against the current basis across the full term.
- Strike the inactive modules before discussing the package price, because 18 to 32 percent of bundled modules were inactive in the agreements reviewed.
- Price the removal path before you need it. The Oracle practice models the metric position before the proposal arrives, not after.
Frequently asked questions
What is the Oracle per employee metric?
A licensing basis that counts your employee population rather than installations or users. The Java SE Universal Subscription is the template, and the definition usually includes contractors, agents and outsourced staff.
Why does the employee metric matter more than the rate?
Because it decouples price from usage. Once headcount is the counting basis, the operational discipline that used to reduce a bill no longer connects to it.
Who owns the risk under per employee pricing?
Corporate development, not IT. It is the only common Oracle metric where a decision taken with nobody from IT in the room changes the licensing bill on the day the deal closes.
Does removing the software reduce the cost?
Not partially. Removing it from most of the estate changes nothing under the employee metric until it is gone from every part of the estate, which is a different project with a different budget.
How does Oracle verify an employee count?
Easily, because headcount is public or readily established. A compliance conversation under this metric needs no estate data and no cooperation from the customer.
What should a buyer do when a metric change is proposed?
Build a counter model before the commercial conversation. Price the same coverage on the current metric across the full term, price the removal path, and test a fifteen percent headcount move in each direction.
What is the core factor and when does it apply?
It applies under Processor licensing, where the count is physical cores multiplied by a factor from the Oracle Processor Core Factor Table. On most x86 platforms that factor is 0.5.
How should a bundle proposal be read?
As a proposal about optionality. A bundle lowers this year's number and reduces the decisions you can take independently later, so the question is which option you are selling and what it is worth.
How much of a typical bundle is inactive?
Between 18 and 32 percent of modules in the agreements reviewed. Striking the inactive ones is a scope conversation, and it belongs before the discussion of package price.
Is a metric change ever the right answer?
Sometimes, where the estate is genuinely everywhere and the counting cost is real. The test is whether the counter model still favors it after growth assumptions and the removal path have both been priced.
