Unused licenses are the quietest line in the software budget and one of the largest pools of recoverable spend. This report reads how much sits idle by category, why it builds up, and the buyer side moves that recover it at the renewal.
Unused software is the quietest line in the enterprise budget and one of the largest pools of recoverable spend. This report reads how much sits idle by category, why it builds up, what it really costs, and the buyer side moves that recover it at the renewal.
About this report. Every number here is a defensible band, not a false point estimate. The proprietary figures come from the Redress Compliance advisory engagement file for 2024 to 2025, cross checked against public vendor pricing and independent spend research.
We say 25 to 40 percent, not 31.4 percent, because real estates differ and precision you cannot defend is worse than an honest range. Where we cite a vendor list price or a usage signal, the source is linked inline so you can verify it yourself.
The audience is the procurement, finance, and software asset owner who has a renewal coming and wants to know where the recoverable money sits before the vendor sends the first quote.
Roughly 25 to 40 percent of paid licenses sit unused or barely used in the average estate we review. That is the headline band, and like every band in this report it hides a wide spread by category and by company.
The figure is not a vendor statistic. It comes from comparing entitlement, the seats and modules a company pays for, against active use measured over a recent window. The gap between the two is shelfware.
Most leaders are surprised by the number the first time they pull it. Spend is tracked carefully. Use is not. The two live in different systems, owned by different teams, and rarely meet on the same page until someone forces them to.
A license is shelfware when it is paid for but not meaningfully used. We separate three states so the number stays honest:
Feature idle is the one buyers miss. A user can log in daily and still never open the analytics module or the advanced security tier that doubled the seat price. The login looks like use. The spend is still wasted.
Active use is not a single number. We read it on three signals and take the strictest one that the platform supports.
A 90 day window is our default for seat products. It is long enough to clear holidays and leave, short enough to catch the genuinely idle. For seasonal teams we widen it. The point is a consistent rule, applied the same way across every contract.
Read the chart as direction, not decimals. Collaboration suites show the smallest gap because almost everyone uses email and chat. AI add on seats show the largest, because they were bought ahead of a use case that has not arrived yet.
Two companies with identical headcount can land 15 points apart on unused share. The driver is governance, not industry. An estate with tight joiner and leaver process and an annual usage review runs lean.
One that buys at renewal and never looks again drifts high. So we report the band and tell you where in it you are likely to sit. If you have never pulled a usage report, assume the top of the range until the data says otherwise.
The most common mistake is sizing the renewal to headcount. Headcount tells you how many people you employ, not how many use a given tool, and the two are rarely the same number.
Sizing to headcount guarantees shelfware in any product that is not used by everyone. The right anchor is active use in the window you chose, not the size of the workforce.
You rarely need a new tool to start. The usage signal is almost always already in the platform you pay for, in the admin console or the identity layer that grants the seats.
Microsoft, Google, Salesforce, and most major suites expose last activity to an administrator. The work is pulling it, putting it next to the entitlement count, and reading the gap, not buying software to discover what you already own.
Zoom out from a single estate and the pattern scales. Independent research has put the share of enterprise software that is bought but underused in the high tens of percent, and our own reviews land in the same broad territory.
As global software spending has kept climbing through 2026, a rising slice of that growth is price and bundling rather than new use. Shelfware is the buyer side shadow of that trend, the paid capacity the spend growth leaves behind.
We do not publish a single market percentage, because the honest figure is a wide band and depends entirely on how use is measured. The direction is not in doubt. The absolute number is a range, not a headline.
Every study counts use differently. Some read logins, some read feature touch, some survey administrators. The methods produce different numbers from the same reality, which is why a responsible figure is always a band.
What every method agrees on is that the share is large enough to matter at the line item level. For a buyer, the market average is interesting. Your own number is what pays the bill.
Shelfware is one kind of software waste, not all of it. Keeping the categories separate matters, because each one is proven with different data and recovered with a different move.
Shelfware versus the waste it is often confused with
| Type of waste | What it is | How you prove it |
|---|---|---|
| Shelfware | Paid seats or modules no one uses | Usage and last active date |
| Wrong tier | An edition richer than the workflow needs | Feature level use against the tier |
| Overlap | Two tools doing the same job | Capability mapping across the estate |
| Over scoped | More capacity than the headcount needs | Entitlement against active users |
Shelfware is the easiest to prove. Usage data is binary enough to settle an argument, which is why it makes the strongest opening case at a renewal. Overlap and wrong tier take more analysis and more internal debate.
When you sit down with a vendor, lead with the waste you can prove cold. A dormant seat report is hard to dispute. An argument about whether you need the premium edition invites a sales counter.
All three belong in a full spend review. But shelfware is the wedge that opens the conversation, because the data is on your side before anyone debates strategy.
Shelfware is rarely one bad decision. It is the slow product of how enterprise software is bought, assigned, and renewed. Four forces do most of the work.
Volume tiers reward bigger commitments, so the discount math nudges buyers to round up. A team that needs 380 seats buys 500 to clear a price break. The extra 120 are shelfware from day one, dressed up as a saving.
Sales incentives push the same way. The account team is measured on committed value, not on whether the seats get used, so the path of least resistance is always a larger number.
People change roles, move teams, and leave. Each event can strand a license that no process reclaims. Over a few years, role drift quietly converts active seats into idle ones across a large workforce.
Offboarding is the worst offender. When a leaver is deactivated in the directory but not removed from the application, the seat keeps billing. We routinely find ghost seats tied to people who left more than a year ago.
Vendors bundle for a reason. A suite that includes modules you did not ask for raises the floor price and trains the buyer to treat the extras as free. They are not free. They are baked into a higher per seat rate.
Independent spend research backs the pattern. As noted where Gartner has tracked rising software spend, a growing share of software cost growth now comes from price and bundling rather than new deployment, which is exactly how padding compounds.
The quietest driver is the renewal that gets waved through. When a contract renews on the same count year after year because no one pulled the usage data, last year mistakes become this year baseline.
Autopilot renewals are why shelfware compounds rather than corrects. Each cycle is a chance to reset the count down. A renewal nobody examined is a chance handed back to the vendor.
Underneath all four drivers sits one root cause. The team that buys the software, the team that assigns it, and the team that pays for it rarely share a single view of who actually uses it.
That information gap is the vendor advantage. They can read your consumption patterns better than you can. Closing the gap on your own side is the precondition for every other move in this report.
The license fee is the visible part. The real cost of an idle seat is the fee plus three quieter charges that grow over the life of the contract.
What a single shelved seat actually costs over a three year term
| Cost element | Visible at purchase | What it becomes by year three |
|---|---|---|
| Base license fee | The sticker price | Carried in full, used by no one |
| Annual uplift | Usually ignored | Compounds on the idle seat every year |
| Audit exposure | Zero | Counts toward entitlement and true up risk |
| Opportunity cost | Zero | Budget that could have funded a needed tier |
This is the charge buyers underestimate most. Vendors apply the annual uplift to your whole base, including the seats no one uses. So an unused license does not hold steady. It costs more every year it stays on the paper.
The chart shows the shape. A seat bought at full rate and never used can carry a meaningful cumulative increase across a three year term, purely from the uplift riding on the idle base. You are paying a rising price to not use something.
Idle seats still count as entitlement, and entitlement is what an audit measures against deployment. A bloated count raises the stakes of any true up dispute and can mask the seats you actually need to defend.
Identity platforms make the idle part visible. Microsoft exposes last sign in and license assignment data, so the data to right size is usually already in your tenant, waiting to be pulled.
Every dollar on a shelved seat is a dollar not spent on a tier someone actually needs. In a flat software budget, shelfware does not just waste money. It crowds out the upgrade or the new capability the business is asking for.
Take a team paying for 1,000 seats of a premium suite where a third sit dormant. That is around 330 idle seats carrying the full rate plus the yearly uplift on each one.
Cut the count to active use at the renewal and you remove the fee and the future increase on those seats in one move. Wait, and you renew the whole 1,000 at a higher rate, then do it again next year. The cost of waiting is the compounding, not just the fee.
Doing nothing is not free. It is often the most expensive option, because the idle base renews at a higher rate every cycle and the count only grows as more drift accumulates.
A renewal waved through on last year numbers locks in last year mistakes and adds this year uplift on top. The cost of inaction is the compounding, paid every year you postpone the reset.
The standard advice is to treat shelfware cleanup as an ongoing hygiene task, a tidy up you do whenever you get around to it. We disagree that hygiene is enough. In the estates we review, shelfware quietly inflates the base that every annual uplift is applied to, so the cost compounds whether or not anyone ever uses the licenses. The buyer side move is to time the cleanup to the renewal, reset the entitlement count down before the uplift is applied, and negotiate swap rights so future drift can be reallocated rather than rebought. Hygiene removes the seat. The renewal removes the seat and the increase riding on it.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Shelfware is not spread evenly. It clusters in the newest and most speculative purchases and thins out in the tools everyone touches every day.
These top the idle ranking. AI add on seats commonly run 40 to 55 percent unused within two quarters, because they are bought on a promise of productivity before the workflows exist.
The Microsoft 365 enterprise plans show how a 30 dollar add on can roughly double a knowledge worker seat, which makes idle AI seats expensive fast. Premium security and analytics tiers follow the same shape. The base product gets used. The expensive upper tier that justified the upgrade often does not.
CRM and sales platforms sit in the middle of the ranking. The Salesforce editions and pricing show how editions step up in price, and buyers frequently land a tier higher than the daily workflow needs.
Consumption products hide shelfware differently. There are no idle seats, but there is committed spend you never draw. An over sized annual commitment that expires unused is shelfware by another name.
The practical lesson is to aim your cleanup at the top of this list first. The recoverable spend per seat is far higher on a premium tier than on a collaboration license, and a lapsed commitment can dwarf both.
The shape repeats across the large suites. The base product gets used. The premium edition, the security upper tier, and the new AI seat are where the idle share concentrates.
That is not a knock on the products. It is a timing problem. The expensive tiers are bought ahead of adoption, and adoption either arrives slowly or never reaches the people who were licensed for it.
Recovery is a sequence, not a one off purge. The aim is to walk into the renewal with evidence, a target count, and the clauses that stop the problem returning.
Start here. Pull last sign in or last active date for every seat and flag anything dormant for 90 days. Most platforms expose this natively, so the first pass takes an afternoon, not a project.
Then add a feature level check on premium tiers. A seat can be active but never touch the tier you paid for, and that idle tier is often the most expensive line to recover.
Removing today shelfware is half the job. The other half is stopping it from rebuilding. A swap or reallocation clause lets you move committed value from a product people abandoned to one they now need.
Without it, every future shift means buying the new product from scratch while still paying for the old one. With it, drift becomes a reassignment. It is one of the most undervalued clauses in any renewal.
Shelfware is not a procurement mistake to apologize for. It is a renewal input to manage, and the renewal is the moment it costs the least to remove.
Vendors discount against evidence, not against complaints. A documented low utilization rate is a stronger argument for a lower count and rate than any plea about budget pressure. Walk in with the usage report open.
The same data also protects you in an audit. A clean, current picture of who uses what is the best defense against a true up claim built on your own bloated entitlement.
Recovery is a small team sport. You need the application administrator who can pull usage, the contract owner who knows the renewal date, and the finance owner who can book the saving.
Keep the group small and the mandate clear. The failure mode is a large committee that debates strategy while the renewal window closes and the count rolls over untouched.
Order matters in the room. Reset the count down to active use first, then negotiate the rate on the smaller base. Doing it the other way lets the vendor discount a number you were never going to keep.
A smaller base also changes the tone. You are no longer asking for a favor on price. You are paying for what you use and negotiating the rate on that, which is a far stronger position.
Cutting the count once is a project. Keeping it cut is a habit. The estates that stay lean treat utilization as a standing number, not a renewal year scramble.
Review usage quarterly, not once every three years when the contract is up. A light quarterly pull catches drift while it is small and keeps the renewal year from becoming a fire drill.
Shelfware grows fastest where no single person owns the number. Assign software asset ownership with a clear mandate to reclaim and reallocate, and the estate stops drifting between renewals.
Ownership does not need a large team. It needs one accountable person, access to the usage data, and the authority to deactivate a seat without a committee. That combination is rarer than it sounds.
Negotiate the right to your own usage data where the platform does not expose it cleanly. If you cannot measure use, you cannot prove shelfware, and the vendor keeps the information advantage at every renewal.
You do not need a dashboard to know the habit is working. A few simple signals tell you the estate is staying lean between renewals.
A recovered saving that is not booked tends to evaporate. When you cut the count, move the money out of the software line in the budget so it cannot quietly refill with new shelfware.
Booking the saving also makes the work visible. A reclaimed seat that shows up as a lower forecast is far easier to defend and repeat than one that disappears back into the same line it came from.
Knowing when each vendor lets you cut a seat is half the timing battle. Most enterprise agreements only allow a reduction at a defined point, and missing that window costs you a full cycle.
When the major vendors let you reduce a seat count
| Vendor or agreement | When you can reduce | The buyer side lever |
|---|---|---|
| Microsoft EA | Generally at the anniversary or renewal, not mid term | Time the true down to the anniversary window |
| Salesforce | At renewal, with no mid term drops | Reset the count from usage before discussing rate |
| Oracle | Support reprices when you terminate part of a base | Model the support repricing before partial termination |
| Adobe ETLA | True forward in term, reset at renewal | Reset the baseline at renewal and resist true forward creep |
| ServiceNow | At renewal on the subscription count | Bring usage data to justify a lower subscription count |
| SAP named user | At renewal, alongside indirect access | Reconcile named users and indirect use in the same pass |
The thread across all of them is the same. Reductions happen at renewal, not on demand, so the work has to be done before the window opens, not after the quote lands.
In most agreements you cannot simply drop seats mid term. The contract commits you to the count for the term, so the renewal is the one moment the number is genuinely up for negotiation.
That is why timing the cleanup to the renewal matters so much. Miss the window and the bloated count rolls into another term, carrying another round of uplift with it.
Several common terms quietly remove your ability to reduce. Read for them before you sign, because each one hands the vendor the count for longer than you intended.
None of these is automatically bad. Each is fine if you chose it with open eyes. The trap is signing one without realizing it locked the count, and the shelfware with it.
The unused share tends to rise with size, but the reason changes as you scale. Small estates pad through bundles. Large estates pad through agreements bought ahead of need and years of role drift.
How shelfware tends to scale with company size
| Company size | Typical unused share | Main driver |
|---|---|---|
| Under 500 staff | 15 to 30% | Bundle padding on entry tiers |
| 500 to 5,000 staff | 25 to 40% | Role drift and offboarding gaps |
| Over 5,000 staff | 30 to 45% | Agreements bought ahead of need |
Larger organizations also carry more ghost seats, because offboarding gaps multiply across more systems and more people. The recoverable spend is material at every scale, but the playbook shifts with the dominant driver.
Whatever the size, the recovery sequence is the same. Measure use, reset the count at the renewal, and put a clause in place so the next cycle starts from a clean base rather than last year mistakes.
Finance does not need to run the usage pull to control the outcome. A short list of questions, asked early, forces the data to surface before the vendor sets the frame.
If the answer to any of these is a shrug, that is the signal to pull the data before the renewal, not after. The questions cost nothing and reframe the whole negotiation.
If a renewal is inside the next twelve months, the work below pays for itself before the first vendor conversation.
In the estates we review, roughly 25 to 40 percent of paid licenses sit unused or barely used. The share runs higher for premium tiers and recent add ons and lower for core collaboration. Treat any single figure with caution. The honest answer is a band, and yours depends on how strictly you count active use.
Shelfware is software a company has paid for but does not meaningfully use. It happens through over buying at the point of sale, role drift after people change jobs or leave, and bundles that include modules no one asked for. None of these is a scandal. They are the normal drift of any large estate left unmanaged between renewals.
Beyond the license fee, unused seats carry the annual uplift, audit exposure, and opportunity cost. The uplift is the quiet one. Vendors apply the yearly increase to your whole base, including the idle part, so an unused seat costs more every year it stays on the contract. The license fee is only the visible tip of the cost.
Start with a usage pull. Most platforms expose last sign in or last active date, so a 90 day inactivity report surfaces the obvious idle seats in an afternoon. Then layer feature level use for premium tiers, since a seat can be active but never touch the tier you paid for. Compare entitlement to active use, not to headcount.
The renewal is the cheapest moment to cut shelfware. At renewal you can reduce the entitlement count before the annual uplift is applied to it, so every seat you remove avoids both the fee and the compounding increase. Mid term cuts are harder because most contracts do not let you drop seats until the term ends.
Premium tiers and recent AI add ons show the worst use in our reviews, often 40 to 55 percent idle within two quarters of purchase. Analytics, business intelligence, and security upper tiers follow. Core collaboration and email sit lowest because almost everyone uses them. The pattern tracks how recently and how speculatively the seats were bought.
Rarely as a cash refund, but often as leverage. Vendors seldom pay you back for idle seats, yet a documented low utilization rate is one of the strongest arguments for a lower renewal count and a better rate. The credit shows up as a smaller bill, not a check. Bring the usage data to the table, not just the request.
Swap rights let you reallocate spend instead of rebuying it. A swap or reallocation clause lets you move committed value from a product people stopped using to one they now need, without buying the second from scratch. It turns future drift into a reassignment rather than a new purchase. It is one of the most undervalued clauses in a renewal.
Yes. Larger estates tend to carry a higher unused share because enterprise agreements are bought ahead of need and role drift compounds over more people. Smaller companies see shelfware mostly from bundle padding on entry tiers. The driver changes with size, but the recoverable spend is material at every scale we benchmark.
No, it is one part of it. Shelfware is paid capacity no one uses. Overspending also includes paying a premium rate for capacity you do use, duplicate tools that overlap, and editions richer than the workflow needs. Shelfware is the easiest part to prove with data, which is why it makes the strongest opening case at a renewal.
The unused share bands by category, the 90 day usage pull method, the swap and reallocation clauses, and the renewal timing that recovers the most before the uplift applies.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement and finance leaders heading into a renewal cycle.
The vendor prices the uplift on your whole base, including the part no one opens. Cancel the idle seats before the renewal and you stop paying to not use them.
