"We have 500 licenses" sounds like a clear statement. It is not. Five hundred licenses counted per user, per device or per processor core describe very different rights, and very different bills. The license metric, the unit a publisher uses to count your use, is one of the most important details in any software contract, and one of the most frequently misunderstood.
This guide explains the common license types, how each is counted, and where each tends to go wrong.
The common license metrics
| Metric | What one license covers | Typical products | Where it goes wrong |
|---|---|---|---|
| Named user | One identified person, on any device | Microsoft 365, Salesforce, Adobe | Licenses stay assigned to leavers and inactive users |
| Per device | One device, used by anyone | Shared kiosks, lab software | Devices retired or replaced without reassigning |
| Concurrent user | One simultaneous user, from a pool | Engineering and design tools | Pools sized for peak, which may happen rarely |
| Per core or processor | Compute capacity of a server | Databases, middleware, virtualization | Virtual machines that can move across hosts expand the count |
| Subscription seat | A named user for a period, usually SaaS | Most SaaS applications | Auto-renewal at the same volume regardless of use |
| Usage-based | Units consumed: calls, storage, credits | Cloud services, AI tools | Unexpected consumption spikes |
| Enterprise-wide | Every employee, or every user in scope | Some Java, analytics and security agreements | Paying for the whole organization when few people use it |
Named user
The most common metric in SaaS. One license belongs to one identified person, who can usually use it on several devices. It is simple to count and simple to waste: when people change roles or leave, their licenses often stay assigned. A reclamation process based on last sign-in recovers most of that waste.
Per device
One license covers one device, regardless of how many people use it. Useful for shared workstations, labs and kiosks. The risk is the reverse of named user: devices get replaced, and the license is not moved to the new device.
Concurrent user
A pool of licenses is shared, and the count is the number of people using the product at the same time. Concurrent pools are efficient but often sized for the busiest day of the year. Usage logs show the real peak.
Per core or per processor
Server software, databases, middleware and virtualization platforms are often licensed by compute capacity. The number of licenses depends on the number of physical cores, sometimes adjusted by a factor that depends on the processor type. Virtualization is the main trap: depending on the publisher's rules, moving a virtual machine across a cluster can require licensing every host it could run on.
Subscription seat
A named-user license for a fixed period, the default for SaaS. It renews, often automatically, and often at the same volume, which is why subscription estates accumulate idle seats.
Usage-based
You pay for what you consume: API calls, storage, compute time or AI credits. There is no compliance gap in the traditional sense, but costs can be hard to predict. Zylo's 2025 SaaS Management Index found that 66.5% of IT leaders had experienced unexpected SaaS charges due to consumption-based or AI pricing.
Enterprise-wide
Some agreements count every employee in the organization, regardless of how many actually use the product. Oracle's Java SE Universal Subscription, introduced in 2023, is a well-known example: it is priced per employee, and the definition of employee includes contractors and consultants who support internal operations. This makes cost easy to calculate and potentially very large compared to actual use.
How metric choice changes the bill
Every metric is a trade-off. Some make the bill easy to predict but can charge you for people who never use the product; others follow real use closely but can surprise you at the end of the month.
Same estate, different metrics
To see how much the metric matters, take one specialist engineering application used by a team of 400 people. Most of them use it a few times a week; at the busiest hour of the busiest day, 150 are using it at once. They work on 560 devices between them.
The concurrent pool is the cheapest here because the peak is well below the headcount. For a tool everyone uses all day, the ranking would reverse. The point is not that one metric is always better, but that the choice should be modeled, not accepted by default.
Questions to ask about any metric
Before you sign, and before every renewal
How MI One helps
Frequently asked questions
Which license metric is cheapest?
It depends on how the software is used. Named-user licensing suits products everyone uses daily; concurrent licensing suits specialist tools used occasionally.
Can we change the metric of an existing contract?
Sometimes, usually at renewal. It is worth modeling the cost under each metric the publisher offers before negotiating.
Are SaaS products always licensed per user?
Most are, but many combine a per-user base with usage-based charges for storage, API calls or AI features.
Sources
- Zylo, "2025 SaaS Management Index," January 16, 2025. https://zylo.com/news/2025-saas-management-index
- Oracle, "Oracle Java SE Universal Subscription Global Price List." https://www.oracle.com/a/ocom/docs/corporate/pricing/java-se-subscription-pricelist-5028356.pdf