Shelfware, the quietest line in the budget and the largest recoverable pool
Unused software is the quietest line in the enterprise budget and one of the largest pools of recoverable spend: roughly 25 to 40 percent of paid seats sit unused in the average estate, and the share is worse for premium and AI tiers. Spend is tracked carefully and use is not, the two living in different systems owned by different teams, and shelfware is a renewal design problem rather than only a procurement mistake, because the unused seats keep paying the annual uplift.
Prepared by Redress Compliance · August 8, 2026 · Cross vendor advisory. Based on 60 to 80 enterprise software spend reviews run or benchmarked 2024 to 2025.
Executive summary
The three states keep the number honest, and feature idle is the one buyers miss.
A license is shelfware when paid for but not meaningfully used, in three states: dormant, assigned to someone with no sign in for 90 days; unassigned, purchased and never allocated; and feature idle, the seat active while the premium tier that doubled its price is never touched.
The login looks like use and the spend is still wasted, which is why the measurement reads three signals, sign in recency, feature touch on the premium modules, and volume drawn on consumption products, taking the strictest the platform supports.
The gap widens exactly where the price rises.
Against entitlement indexed to 100, active use in the last 90 days ran 72 percent on collaboration suites, 55 on CRM and sales seats, 46 on premium security tiers, 38 on analytics, and just 29 percent on AI add on seats: the premium and AI tiers showed the worst use.
Often 40 to 55 percent idle within two quarters of purchase, bought ahead of a use case that has not arrived.
The cost of the idle seat is the fee plus the compounding uplift plus the audit exposure, never just the sticker.
The renewal is the cheapest moment to cut, and right sizing first paid twice.
The renewal resets the count before the uplift is applied, and right sizing at the renewal rather than mid term recovered 15 to 30 percent of the line before any rate negotiation began, with the buyers who right sized before negotiating rate realizing 40 to 60 percent of the vendor opening ask.
The headcount trap guarantees the opposite: sizing the renewal to headcount rather than active use builds shelfware into any product not used by everyone, and the right anchor is measured use in the window you chose.
The data already exists, and the governance is what varies.
The usage signal is almost always in the platform you pay for, the admin console or identity layer exposing last activity.
And the work is pulling it next to the entitlement, which two companies with identical headcount can differ on by 15 points of unused share purely through governance: joiner and leaver process, and an annual usage review. If you have never pulled a usage report.
Assume the top of the range until the data says otherwise, and negotiate swap and reallocation rights so future drift moves rather than rebuys.
Entitlement against use, by category
| Category | Active in the last 90 days | The pattern |
|---|---|---|
| Collaboration suites | 72 percent | The smallest gap, since almost everyone uses email and chat |
| CRM and sales seats | 55 percent | Role drift and leavers accumulating in the pool |
| Premium security tier | 46 percent | The feature idle classic: seats active, tier untouched |
| Analytics and BI | 38 percent | Bought broadly, used by the few who analyze |
| AI add on seats | 29 percent | Bought ahead of a use case that has not arrived |
Read the chart as direction, not decimals, and read your own estate against it.
The 90 day window is the default for seat products, long enough to clear holidays and leave, short enough to catch the genuinely idle, widened for seasonal teams, and the point is a consistent rule applied identically across every contract.
The band is a band because governance, not industry, drives the spread: the estate with tight joiner and leaver process and an annual usage review runs lean, and the estate that buys at renewal and never looks again drifts high.
Finding it and recovering it, at the renewal
- Pull the 90 day usage report from the admin console or identity layer, the signal that already exists in every major suite.
- Run the feature level check on premium tiers, because the feature idle state hides behind daily logins.
- Compare entitlement to active use on one page, the meeting the two systems never have until someone forces it.
- Cut the count at the renewal, not mid term, since the renewal resets the base before the uplift applies and mid term reductions rarely process.
- Negotiate swap and reallocation rights, so the next round of drift moves between products instead of being rebought.
The renewal negotiation timing playbook
The right sizing pass, the T minus 12 calendar, and the clause set that keeps the recovered count from rebuilding.
Get the white paper →Why it accumulates, and how to stop it returning
Shelfware rebuilds through the same mechanisms that built it: the headcount sized renewal that ignores active use, the departmental rollout that attaches by org chart, the leaver whose license nobody reclaims.
And the premium tier bundled as a sweetener that nobody measures, which is why the durable fix is cadence rather than a one time sweep, the quarterly usage pull, the joiner and leaver discipline, and the annual review before every renewal snapshot.
The vendor by vendor sweeps run through the practice, the Microsoft method in the usage audit guide, the Salesforce idle base in the Salesforce licensing guide, and the AI attach discipline where the 29 percent lives in the AI ROI report.
The wider pattern family the shelfware line belongs to sits in the overpayment patterns report.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across spend reviews, 2024 to 2025
Across roughly 60 to 80 enterprise software spend reviews we ran or benchmarked between 2024 and 2025, unused entitlement was the single most common source of recoverable cost:
Within two quarters of purchase, the tiers bought ahead of the use case.
Of the line, from right sizing at the reset moment before any rate conversation.
The report's framing for finance is the full cost of the idle seat: the annual fee, the compounding uplift applied to it at every renewal, and the audit exposure of entitlements nobody reconciles, which together make the dormant license more expensive every year it survives.
Most leaders are surprised by the number the first time they pull it, because spend and use live in different systems owned by different teams.
And the entire discipline reduces to forcing them onto the same page on a schedule, with the renewal calendar as the enforcement mechanism and the swap right as the insurance.
Your first five moves
- Pull the 90 day usage report today, and assume the top of the 25 to 40 percent range until the data says otherwise.
- Check feature touch on every premium tier, the idle state that hides behind daily logins.
- Anchor the renewal to active use, never headcount, the trap that guarantees shelfware by construction.
- Time the cut to the renewal, where 15 to 30 percent of the line recovered before rate talk began.
- Write swap and reallocation rights into the order, so drift moves instead of rebuying. The cost optimization practice runs the sweep with you.
Frequently asked questions
How much enterprise software goes unused?
Roughly 25 to 40 percent of paid seats in the average estate, measured by comparing entitlement against active use over a 90 day window, with the share worse on premium and AI tiers: premium security ran 46 percent active, analytics 38, and AI add on seats just 29 percent against entitlement.
Governance, not industry, drives where in the band an estate sits.
What counts as shelfware?
Three states: dormant licenses assigned to people with no sign in for 90 days or more, unassigned licenses purchased and sitting in the pool, and feature idle seats where the user logs in daily but never touches the premium tier that doubled the price.
Feature idle is the state buyers miss, because the login looks like use while the premium spend is wasted.
When is the best time to cut unused licenses?
At the renewal: it is the cheapest moment because the entitlement count resets before the annual uplift is applied, and right sizing at the renewal rather than mid term recovered 15 to 30 percent of the line before any rate negotiation.
Mid term reductions rarely process under standard paper, which is why the usage review belongs on the renewal calendar, months before the snapshot.
What does an unused license actually cost?
More than the sticker: the annual fee, plus the compounding uplift applied to it at every renewal, plus the audit exposure of entitlements nobody reconciles.
A dormant seat gets more expensive every year it survives, and premium tiers bought as sweeteners compound worst because their base price is highest and their measured use, 40 to 55 percent idle within two quarters, is lowest.
How do you measure software usage without new tools?
The signal already exists in the platforms you pay for: Microsoft, Google, Salesforce, and most major suites expose last activity through the admin console or identity layer, and the work is pulling it next to the entitlement on one page.
Read three signals, sign in recency, feature touch on premium modules, and volume on consumption products, applying the same 90 day rule across every contract.
How do you stop shelfware from coming back?
With cadence and clauses: the quarterly usage pull, the joiner and leaver reclaim discipline, the annual review before every renewal snapshot, and swap and reallocation rights written into the order so future drift moves between products instead of being rebought.
The renewal sized to active use rather than headcount is the structural fix, because headcount sizing rebuilds the shelfware every cycle.
