Close to half the estates were sitting on a metric that overcharged the workload
The discount is negotiated once and quoted everywhere. The metric is chosen once, rarely revisited, and decides what the discount is applied to for the life of the estate.
Prepared by Redress Compliance · August 17, 2026 · Oracle advisory. 40 to 50 Oracle licence metric reviews, 2024 to 2025.
Executive summary
Close to half the estates were on a metric that overcharged the workload. Not a rounding issue. A structural mismatch between how the system is used and how it is counted, carried forward unexamined.
Named User Plus minimums forced counts 25 to 40 percent above the real user population. On low user, high core systems, where the per processor minimum has nothing to do with how many people actually use the program.
Named User Plus counts more than users. Every individual authorised to use the program, plus every non human operated device that can reach it, whether or not anyone logged in this quarter.
Support runs at 22 percent of licence value. Which means a metric error is not a one off overpayment. It is an annuity, and it compounds with every renewal.
What each metric actually counts
The two principal metrics measure different things, and the choice between them is a statement about the shape of the workload rather than about price.
| Metric | What it counts | Where it fits |
|---|---|---|
| Processor | Cores, adjusted by the core factor | High user or unbounded populations |
| Named User Plus | Every individual authorised, plus every non human operated device that can reach the program | Genuinely small, countable, stable populations |
| NUP minimums | A floor per processor, regardless of the real population | Nowhere, when the system is low user and high core |
| The support line | 22 percent of licence value, annually | Everything, which is why the metric choice recurs |
The Named User Plus definition is wider than most estates assume, and the width is where the surprises live. It is not a count of people who log in. It counts every individual authorised to use the program, whether or not they used it this quarter, plus every non human operated device that can reach it. An interface, a scheduler, a monitoring agent, and a batch process are all countable under that definition. An estate that sized NUP against active human users has sized it against the wrong population.
The metric outlives the discount, and it is chosen once
Across roughly 40 to 50 Oracle licence metric reviews run in 2024 and 2025, close to half the estates were sitting on a metric that overcharged the workload. That is a strikingly high rate for something that is, on its face, a straightforward decision. The explanation is that the metric is chosen once, usually at the original purchase, against a workload profile that has since changed, and nothing in the ordinary lifecycle of an Oracle estate prompts anyone to revisit it. The discount gets renegotiated. The thing the discount is applied to does not.
The clearest form of the mismatch is Named User Plus on low user, high core systems, where NUP minimums forced counts 25 to 40 percent above the real user population. The minimum is a floor set per processor. On a system with many cores and few users, that floor has no relationship at all to how many people use the program, and the estate ends up paying for a population that exists only in the licensing arithmetic. Named User Plus is designed for genuinely small, countable, stable populations, and a high core system is usually the opposite of that even when the headcount looks modest.
There is a second trap inside the same metric, which is the breadth of the definition. Named User Plus counts every individual authorised to use the program, plus every non human operated device that can reach it, whether or not anyone logged in this quarter. Authorisation rather than use is the test, and devices count. An estate that sized its NUP position against active human users has undercounted against the contract and is carrying compliance exposure alongside whatever it is overpaying on the minimum, which is an unusually poor combination.
What makes the metric worth more attention than the discount is that support runs at 22 percent of licence value. A metric error is therefore not a one off overpayment that washes through; it is an annuity, charged every year and carried into every renewal, and it grows with the estate. The practical move is to test the metric against the workload before the next purchase or renewal rather than after, model both metrics side by side on the systems where the population is small relative to the cores, and count NUP against the contractual definition rather than the login list. The leverage question sits in benchmarks and leverage, the audit exposure in the audit brief, and the library in the Oracle practice.
- Your agreements decoded into plain English before the auditor interprets them for you
- Entitlements, caps, and protections verified across your whole contract portfolio
- A defensible position paper generated in minutes, not weeks
Testing the metric
- Model both metrics side by side on every system where the population is small relative to the core count, which is exactly where NUP minimums bite hardest.
- Count Named User Plus against the contractual definition, every individual authorised plus every non human operated device that can reach the program, not the login list.
- Test the metric before the next purchase or renewal, because it is chosen once and nothing in the ordinary lifecycle prompts a review.
- Treat a metric error as an annuity, since support runs at 22 percent of licence value and the error is charged every year rather than once.
- Re-examine the metric whenever the workload profile changes, which is the event that turns a correct original choice into a wrong current one.
- Negotiate the metric before the discount, because the discount is a percentage of whatever the metric produces.
What the metric reviews showed, 2024 to 2025
Across roughly 40 to 50 Oracle licence metric reviews:
Close to half the estates were sitting on a metric that overcharged the workload, a structural mismatch rather than a rounding issue.
How far Named User Plus minimums forced counts above the real user population on low user, high core systems.
Named User Plus counts every individual authorised to use the program, plus every non human operated device that can reach it, whether or not anyone logged in this quarter. Authorisation rather than use is the test, and devices count.
Support sits at 22 percent of licence value, which turns a metric error from a one off overpayment into an annual charge carried into every renewal.
Watch the briefing · 4:17The Oracle Renewal ConversationWhy the base the discount is applied to matters more than the percentage.
Your first five moves
- List every system where the user population is small relative to the core count, which is where the mismatch concentrates.
- Model processor against Named User Plus on each of them, including the per processor minimum rather than the headline rate.
- Rebuild the NUP count against the contractual definition, authorised individuals plus non human operated devices.
- Price the difference over the support line at 22 percent, so the annual cost of the error is visible rather than the one off.
- Settle the metric before the discount. The Oracle practice runs the comparison with you.
Frequently asked questions
How often is the metric wrong?
Close to half the estates across the 40 to 50 metric reviews were sitting on a metric that overcharged the workload. It is a structural mismatch rather than a rounding issue.
Why does it happen so often?
Because the metric is chosen once, at the original purchase, against a workload profile that has since changed. The discount gets renegotiated at every renewal; the thing the discount is applied to does not.
What are NUP minimums?
A floor on the Named User Plus count set per processor. On low user, high core systems that floor forced counts 25 to 40 percent above the real user population, because it has no relationship to how many people use the program.
What does Named User Plus actually count?
Every individual authorised to use the program, whether or not they logged in this quarter, plus every non human operated device that can reach it. Authorisation rather than use is the test, and devices count.
Which devices count?
Non human operated devices that can reach the program. Interfaces, schedulers, monitoring agents, and batch processes all fall within that definition, which is wider than most estates assume when they size a NUP position.
What is the risk of sizing NUP against logins?
You undercount against the contract, so you carry compliance exposure at the same time as overpaying on the minimum. That combination is unusually poor and it is common.
Why does the support line matter here?
Because support runs at 22 percent of licence value. A metric error is therefore an annuity charged every year and carried into every renewal, not a one off overpayment that washes through.
When should the metric be tested?
Before the next purchase or renewal, not after. Nothing in the ordinary lifecycle of an Oracle estate prompts a metric review, so it has to be scheduled deliberately.
What triggers a re-examination?
A change in the workload profile. That is the event that turns a correct original choice into a wrong current one, and it is exactly the event nobody connects to licensing.
Should the metric or the discount be negotiated first?
The metric. The discount is a percentage of whatever the metric produces, so settling the percentage against the wrong base fixes the error in at a slightly better rate.
