Unused vs underused licences: where the software waste actually sits now
The cheap waste — seats nobody ever opened — is shrinking. The expensive waste, licences people do open and use at a fraction of what you bought, is growing. Vertice’s Q2 2026 data shows the split moving, and it decides which half of your software bill a reclamation script will never find.
Most of the money you waste on software is no longer sitting in licences nobody opened. It is sitting in licences people do open — and use at a fraction of what you bought. That distinction decides which tools you should cancel, which you should teach, and why a licence-reclamation script keeps returning a smaller number than your invoice implies.
How much money do companies waste on unused software licences?
Around two-thirds of SaaS licences are either fully unused or underused. Vertice’s Q2 2026 tracked-spend data puts it at 65% — up from 62% a year earlier, but down from a 69% peak in December 2025. For a company of 500–1,000 employees, the same dataset puts the annual cost at $1.75 million (the figures are US dollars, not euros).
Three caveats that number deserves before it travels any further:
- Vertice sells SaaS procurement and management. This is a vendor measuring the problem it sells the cure for. The finding is still useful; the interest should be named.
- The disclosed methodology is one line: "over $75bn of global processed spend managed by Vertice in 2026." No sample size, no company count, no independent audit.
- The total is not climbing in a straight line. It is up year-on-year and down from December. Anyone telling you "software waste is getting worse" is reading one comparison and ignoring the other.
Which is why the headline percentage is not the interesting part of this data. The composition is.
What is the difference between an unused licence and an underused one?
An unused licence is a seat that has been provisioned, paid for, and never opened — shelfware, a ghost licence, a seat belonging to someone who left. An underused licence is a seat somebody signs into regularly and uses at a fraction of what it can do: the premium tier bought for a capability nobody was ever shown, the platform used as a file store, the AI assistant used to reformat text.
Over the past year those two categories moved in opposite directions. Fully unused licences fell from 15% to 14%. Underused licences rose from 47% to 51% — and are now the clear majority of the waste. (Vertice counts a tool as underused when more than half its purchased seats sit idle, so the real figure is likely higher still.)
One of those numbers is going down because tooling and finance discipline got better at it. The other is going up, and nothing in the standard toolkit is pointed at it.
Why does a licence-reclamation script only find the cheap half?
Because a script enumerates seats, and the second kind of waste is not a seat problem. Last sign-in date, assigned versus unassigned, licence SKU, provisioning status — a PowerShell or Graph API script reads all of it well, and it answers one question accurately: *who has not opened this?*
It cannot answer the question the money now depends on: *who opens this every day and gets a tenth of what we paid for?* That is not a flaw in anyone’s script. It is the ceiling of the instrument. Two more limits come with it: a script cannot separate a seat that is idle from a seat that is strategic-but-dormant, and nobody maintains it once the person who wrote it moves on.
The script is the easy 20%. The adoption is the 80% that actually pays back.
Can a SaaS management dashboard tell you what to cancel?
No. A licence dashboard reports state; cancelling is a decision. Those are different jobs, and only one of them can be automated.
A good dashboard will tell you a seat has been dormant for 90 days. It cannot tell you that the seat belongs to the one person who runs the quarterly consolidation. It cannot choose between two overlapping tools that two departments bought separately, because that choice depends on which team’s workflow you are willing to change. And it cannot make anybody use the tool you decide to keep — which is the only step that converts the remaining spend into a return.
This is a limit of the instrument, not a criticism of the vendors. Plenty of organisations run a capable SAM platform and still cannot produce the one artifact their CFO asks for: a defensible utilisation number with a euro figure attached.
How do you find the underused half?
You measure capability against purchase, tool by tool, and you attach a euro figure to the gap. Concretely, week one of a scoped audit produces three things:
- An inventory of what is actually owned — not what procurement believes is owned. Shadow subscriptions on personal cards are normal, not exceptional.
- A real utilisation number per licence and per seat, produced independently of the vendor whose renewal is at stake.
- A hard € figure, each line traced to an invoice, so the number survives contact with a board rather than starting an argument.
Then there are two recovery categories most teams skip, and they are worth naming because they take nothing away from anyone:
- Downgrade over-licensed users. People sitting on a premium tier using only what the tier below already includes. Same tool, same access to the work they actually do, smaller line item.
- Redefine the licence terms. Commitment length and billing frequency are negotiable inputs, not fixed properties of the contract.
Neither of those requires you to take a tool off anyone’s desk — which matters, because the fastest way to lose a room is to arrive looking like you came to cut their stack.
What this changes about your next renewal
- Split the licence report in two before you read it. Never opened, and opened-but-shallow. They have different owners, different fixes and different arithmetic.
- For every tool, write down what it was bought to do and what it is used for. The gap between those two sentences is the second list, and it is the one that is growing.
- Price the untapped capability, not just the idle seat. Cancelling is one-time money. Adoption is the recurring return.
- Give the second list an owner. A cancellation always has one — someone signs it. A capability nobody uses rarely does, which is exactly why it survives every review.
We do not sell licences and have nothing to upsell, which is why we can tell you to cancel something. What we recover on the first list is what pays for the work on the second.
A script can find an idle seat. It cannot find an untapped person.