The Shelfware Audit: Finding the Subscriptions You’re Paying For and Not Using

The short answer: Shelfware is software you pay for and nobody uses — dead seats from departed employees, duplicate tools, annual plans that renewed on autopilot, AI add-ons nobody opened. You don’t need a study to find your number: pull 12 months of card and bank statements, list every recurring software charge, match each one to an owner and a login count, kill the orphans, downgrade the underused, and put every renewal on a calendar. One afternoon finds the bleed; thirty minutes a quarter keeps it from coming back.

This guide is educational, not financial advice. It’s an operational procedure — every dollar you act on should come from your own statements, not an industry average. The enterprise studies cited below tell you this is a widespread problem; your audit tells you your number.

---

Shelfware, defined: how a small business collects dead subscriptions

Nobody sets out to waste money on software. Shelfware accumulates through four quiet mechanisms:

Ex-employee seats. Someone quits in March; their $25/seat subscription keeps billing through the next three annual renewals because nobody reassigns or cancels the license. Per-seat tools make this automatic — the seat exists whether or not anyone sits in it.

Duplicate tools. The office pays for a survey product, a forms product, and a poll feature inside another product — all doing roughly the same job, bought by three different people at three different times, each on its own card.

“We might need it” annual plans. The annual discount looked attractive at signup, so you pre-paid a year for a tool the team tried twice. The annual plan then auto-renews — at full list, without the intro discount — and the renewal notice goes to the card owner’s email while you’re busy running the business.

AI add-ons nobody opened. Per-seat AI add-ons — Microsoft’s Copilot and OpenAI’s ChatGPT Business run roughly $21 to $30 per seat per month (Copilot Business is $21/user/month on an annual commitment as of July 2026; ChatGPT Business is $25/seat/month) — are the current classic. The seat was bought for a productivity pilot, three people opened it twice, and it’s been billing for everyone ever since. Zylo’s 2026 data finds ChatGPT is now the single most-expensed application in the enterprise sample, which is exactly the adoption pattern that produces this kind of waste.

The common thread: none of this happens because anyone was reckless. It happens because nobody owns the subscription list. Zylo’s 2026 index survey found business units now control 81% of SaaS spend while IT directly manages just 15% — and that’s in companies that have IT departments. In a 15-person business, the “IT department” is whoever clicked “upgrade” last.

What the industry numbers say (and what they don’t)

Here’s the honest frame, because this guide’s value is the procedure, not the numbers:

• Zylo’s 2026 SaaS Management Index (announced July 2026) finds organizations leave an average of 36% of SaaS licenses unused, measured against recommended utilization levels. Utilization was 54% in 2025, up from 47% in 2024 — improvement, and still nearly half the seats unopened.
• Gartner’s SaaS Sprawl report (October 2024) projects that organizations failing to centrally manage SaaS lifecycles will overspend on SaaS by at least 25% through 2027, driven by unused entitlements and overlapping tools.

Two caveats, stated plainly:

1. These are enterprise studies. Zylo’s data comes from its own platform telemetry and a survey of 218 IT leaders — companies with hundreds of applications and dedicated software-asset staff. Your 12-person shop is not their sample. Treat every percentage as directional: it tells you the problem is real and widespread, not what your waste percentage is.
2. The 36% and the 54% are measured differently. “36% unused” is measured against recommended utilization levels (a stricter bar); “54% utilization” is licenses actually used. Don’t subtract one from the other to manufacture a trend — they’re different rulers.

The numbers matter only as motivation. The audit is what gives you your number, and your number is the only one that can save you money.

The repeat offenders, ranked

In small-business audits, the same four charges show up most often. Check for them in this order:

1. Ex-employee seats. The single most common find. Pull your seat counts in each per-seat tool and compare them to your current payroll list. Every mismatch is a candidate for cancellation — and note the trap: some tools bill monthly but only decrement at the annual renewal, so flag the renewal date too.
2. Duplicate tools. Two survey tools, three cloud-storage subscriptions, a video-conferencing add-on on top of a suite that already includes one. Pick the one that fits your workflows and kill the rest.
3. Annual plans nobody opened. Any tool billed annually that shows zero or near-zero logins for months. These are the charges most likely to surprise you, because annual charges only appear on statements twelve times a year — easy to skim past eleven of them.
4. Per-seat AI add-ons. Bought enthusiastically, used by almost nobody. Check actual usage before you renew: if two of ten seats are active, you’re paying 5x per active user.

The afternoon audit: the procedure, step by step

This is the whole guide in practice. Four hours, once.

Step 1: Pull 12 months of statements. Export CSVs from every business card and bank account that could carry a software charge. Filter for recurring charges. Twelve months matters because annual plans hide in a single month — a 90-day lookback will miss them entirely.

Step 2: Build the list. A spreadsheet works. You need four columns: tool name, what it costs and how often, who owns it, last meaningful use. Do not buy a SaaS-management platform to do this — at small-business scale, the spreadsheet is the tool, and the enterprise spend-management products (Zylo, BetterCloud, and their competitors) are priced for companies managing hundreds of applications. That’s a poor fit for your business and a purchase this guide has no interest in selling you.

Step 3: Match each charge to an owner and a login count. For each line item, answer two questions: who is responsible for this tool? and when did someone last log in? Admin consoles in most tools show seat counts and last-login dates — check the three expensive ones first (office suite, CRM, project tools). Anything with no owner and no recent login is shelfware by definition.

Step 4: Kill the orphans, downgrade the underused. Apply the decision rules in the next section. Note the sequence: cancel first, downgrade second, keep third. The temptation is to research alternatives for everything — resist it. You’re removing spend, not redesigning your stack.

Step 5: Calendar every renewal. For everything you keep, record the renewal date and the non-renewal notice deadline — the 30/60/90-day window buried in the contract that determines whether you can exit before the next term locks. This is the mechanical heart of the audit, and it’s where the audit connects to the auto-renewal traps guide: a renewal you know about is negotiable; a renewal you discover on your statement is not.

Step 6: Check the hidden drawers. A few places software charges hide that statements won’t catch on their own: employees expensing personal-card AI subscriptions (ChatGPT is now the most-expensed app in Zylo’s sample — some of yours may be living on expense reports); app-marketplace add-ons billed through a platform (POS and accounting app stores charge $0–$99+/app/month on top of the base subscription); and trial-to-paid conversions that arrived mid-year.

The decision rules: kill, downgrade, keep

Three rules, no sentimentality:

• No login in 60 days → cancel. If nobody has opened it in two months, the tool isn’t serving anyone. The “we might need it again” instinct is the shelfware instinct — kill it, and re-buy later if a real need appears. Re-buying costs nothing; the subscription you keep “just in case” costs real money every month.
• Used by one or two people occasionally → downgrade or drop seats. Check whether the tool offers a lower tier, fewer seats, or a free tier that covers the actual usage. Ex-employee seats fall here: remove the seats, don’t just disable the users.
• Duplicate tools → pick one, kill the rest. Choose based on what the team actually uses, not the feature list — the feature list is how you got two tools in the first place.

What does this look like in dollars? An illustrative pass through a typical 12-person team’s findings: three Copilot seats nobody opened ($21 × 3 × 12 = $756/year), four ex-employee seats on a project tool ($25 × 4 × 12 = $1,200/year), a survey product duplicating a feature already in the office suite ($49 × 12 = $588/year), and one annual plan that auto-renewed twice unnoticed ($399/year). Total: $2,943 a year — found in an afternoon, with nothing the team was actually using touched.

Keeping it clean: thirty minutes a quarter

The audit decays. Every new signup is a future orphan. The maintenance routine:

• Offboard the seats. Employee exits should include a license step: reassign or cancel every per-seat license on the day of departure, not “when someone remembers.” This one habit prevents the #1 shelfware source.
• Quarterly seat review. Fifteen minutes: open each per-seat admin console, compare seat counts to payroll, drop the mismatches.
• Renewal calendar review. Check the calendar you built in Step 5 before each renewal’s notice deadline. The contract tells you when you can leave; the calendar makes you remember to check. (SaaS contract red flags for what to look for in those terms; per-seat vs. usage pricing for when the renewal makes you question the model itself.)
• AI add-on review. Re-verify per-seat AI usage every quarter — adoption either justifies the seats or the pilot is over. At $21–30/seat/month, this is the fastest-growing line item to audit, and the one most often bought on enthusiasm rather than measured use.

Methodology

Industry statistics are attributed with vintage and sample caveat, not asserted as small-business facts: the 36%-unused figure and 54% utilization are from Zylo’s 2026 SaaS Management Index (announced July 2026; platform telemetry plus a survey of 218 IT leaders — an enterprise sample); the “at least 25% overspend through 2027” projection is Gartner’s SaaS Sprawl report (October 2024), quoted via 2026 secondary sources since the original is an analyst report, not a public page. The 81%-business-unit/15%-IT spend-control split is from the same Zylo 2026 survey. AI add-on pricing ($21/user/month Copilot Business on annual commitment, post–July 2026 cut; $25/seat/month ChatGPT Business) is from July–October 2026 pricing reporting — verify at time of purchase, as this category reprices frequently. The dollar-worked example in “The decision rules” is illustrative arithmetic on the listed per-seat prices, not a verified case. No anecdote in this guide is presented as a real company’s audit.

Sources

• Zylo, “Announcing Zylo’s 2026 SaaS Management Index” — zylo.com (36% of licenses unused against recommended utilization levels; survey of 218 IT leaders; 81% of spend business-unit-controlled), announced July 2026
• UseCarly, “SaaS Statistics 2026: App Sprawl, Spend, and Waste” — usecarly.com (compilation: 54% utilization in 2025, up from 47% in 2024; Gartner 25% projection context), accessed October 2026
• Gartner, SaaS Sprawl report (October 2024) — organizations failing to centrally manage SaaS lifecycles will overspend by at least 25% through 2027; quoted in CloudEagle, cloudeagle.ai, and BetterCloud’s 2026 ROI-guarantee announcement, accessed October 2026 (original report is analyst-gated; quote via secondary sources)
• Zylo press release on the 2026 index (ChatGPT now the most expensed app; expense-based SaaS spend up 267% YoY) — syndicated PR, e.g. kynt1450.com, July 2026
• SeatCompress, “Microsoft 365 Copilot Dropped to $21” — seatcompress.com ($21/user/month Copilot Business on annual commitment, cut July 1, 2026; $25.20 month-to-month; $18 promotional rate through 2026), August 2026
• AI Pricing Guru, ChatGPT Business vs. Microsoft 365 Copilot Business — aipricing.guru (ChatGPT Business $25/seat/month; $20 annual), updated October 2026

Frequently asked questions

How often should I run the full audit? Once a year, minimum. The full 12-month statement pull is what catches annual plans; anything shorter misses them. The quarterly maintenance (seat review, renewal check) is what keeps the annual audit small.

Who should own this in a 10-person business? Whoever controls the cards. That’s usually the owner or the bookkeeper. The failure mode is “everyone bought something, nobody owns the list” — so make one person the list-owner, and have them sign off on the quarterly review.

Does the audit replace auto-renewal management? No — it’s the discovery step. The audit finds what you’re paying for; the auto-renewal guide covers how renewals extract money and how to defend against them. Run the audit first, then manage what survives it.

What about free trials? Trials are the pre-shelfware stage. California’s amended auto-renewal law now explicitly covers free trials converting to paid — but the practical defense is the same: calendar the conversion date at signup, or you’ll own a subscription you never decided to buy.

My team is five people. Is this overkill? A five-person team can still bleed $200–400/month on dead seats and duplicates — real money at that size. The audit just takes an hour instead of an afternoon. The procedure scales down fine.

Should I just buy a SaaS-management tool to do this? For a small business, no. Those tools are priced for enterprises managing hundreds of applications, and a spreadsheet does the job at your scale. Spend the audit money on canceling subscriptions, not on software that finds them.