SaaS Sprawl: Audit What You Already Pay For Before You Switch Anything

Zylo's 2026 benchmark, built on more than 40 million licenses, found organizations leave an average of 36% of their SaaS licenses unused — and that business units, not IT, control 81% of software spend. Both numbers break the arithmetic of a switching decision before you start it. Here's the audit that comes first.

Why the audit comes before the comparison

A switching-cost calculation has a simple shape: what you pay now, what you'd pay after, and what the move itself costs in between. Every part of that depends on knowing what you currently pay — and in most organizations that figure is both wrong and knowable only with effort.

The 2026 SaaS Management Index from Zylo, a SaaS management vendor, is in its eighth year and draws on more than 40 million SaaS licenses and $75 billion in spend under management. (It is a vendor report with a commercial interest in the problem it measures, so treat the direction as more reliable than the decimal place.) Three of its findings matter directly for anyone pricing a switch:

• 36% of licenses go unused, measured against industry-recommended utilization levels. Your "current cost" line is partly a line for seats nobody opens.
• Business units control 81% of SaaS spend; IT directly manages 15%. Whoever is running the comparison probably cannot see most of the stack from where they sit.
• Expense-based SaaS spend rose 267% year over year, with ChatGPT the most-expensed application. A growing share of the stack arrives on employee cards and never touches procurement at all.

None of this says "don't switch." It says that the baseline you're switching away from is usually inflated, invisible in parts, and changing underneath you. Zylo also reports that large enterprises add an average of 21 applications a month even while total app counts look flat — the portfolio is churning faster than it is growing.

How big is the waste, really? It depends what you count

Published waste figures range from roughly a quarter to more than half, and the spread is mostly definitional rather than a disagreement about facts. "Licenses that are completely idle" is a different measurement from "licenses used below a recommended threshold," which is different again from "share of the IT budget spent on redundant or underused software." It's worth knowing which one a number refers to before quoting it in a business case.

Industry guidance commonly suggests that reclaiming idle licenses and rightsizing over-specified tiers can cut SaaS spend by something in the range of a quarter to 40%. That is a consultancy-and-vendor estimate rather than audited research, and the achievable number in any one company depends entirely on how much slack is actually there. But the order of magnitude matters, because it is often the same order of magnitude as the saving a proposed switch is supposed to deliver.

Three ways sprawl changes the switching math

1. The cheaper option may already be in a contract you hold. Redundancy is the most common finding in a first audit: two project trackers, three video tools, overlapping storage. When the "new" tool duplicates a module inside a suite you already pay for, the comparison isn't tool A versus tool B — it's tool A versus a licence you've already bought. Our Slack-to-Teams worked example is the clearest case of this: Teams looks free precisely because Microsoft 365 is already on the invoice, which is also why that comparison is easy to get wrong in the optimistic direction.

2. Your incumbent is probably cheaper than your invoice says. If a third of your seats are idle, the honest comparison is between the incumbent after rightsizing and the challenger at the seat count you'd actually provision. Many switches that pencil out against a bloated baseline stop penciling out against a trimmed one. Running the reclaim first is also strictly lower-risk: no migration, no retraining, no dual-running period.

3. Rightsizing and switching compete for the same renewal window. Both depend on the same leverage, and that leverage has a short life. Usage data showing you're at 60% utilization is the strongest input to either conversation — a reason to downsize the renewal, or evidence that the tool was never adopted and should go. Starting that work 90 or more days out is the practical difference between negotiating and accepting; see our note on renewal timing and what to ask for.

A minimum viable audit

Full SaaS management platforms exist and may be worth it at scale. For a first pass, most of the signal is available from systems you already run:

1. Pull the financial record first. Accounts-payable vendor lists, corporate card statements, and expense-report line items. Given that expense-based SaaS spend is the fastest-growing intake channel, this is where the surprises are — not in the procurement system.
2. Pull the identity record second. SSO and identity-provider logs show which applications are actually being signed into, and by how many distinct people. The gap between "licenses purchased" and "distinct monthly sign-ins" is your rightsizing estimate in its roughest form.
3. Reconcile the two. Applications in the financial record but not the identity record are either unmanaged or abandoned. Applications in the identity record but not the financial record are free tiers that may be holding business data without a contract behind them.
4. Name an owner for every line. With business units controlling most spend, an unowned application is one nobody will cancel and nobody will defend. Ownership is what makes the next step possible at all.
5. Build a renewal calendar. Sort every contract by renewal date and auto-renewal notice window. This single artifact converts the audit from a report into a sequence of dated decisions — and it's the one that tends to survive after the audit is over.
6. Only then run the switching comparison, using the rightsized incumbent cost as the baseline rather than the current invoice.

Treat the output as three buckets, not one: cancel (no owner, no usage), rightsize (real usage, wrong seat count or tier), and replace (real usage, wrong tool). Only the third bucket is a switching decision, and it is usually the smallest of the three. The first two cost nothing but attention. The third is the one that needs the full total-cost-of-ownership treatment, including migration labor and the overlap period when both vendors are billing.

The AI complication

AI tooling is where sprawl is growing fastest and where the audit is hardest. Zylo reports AI-native application spend up 108% year over year overall and 393% in organizations above 10,000 employees, arriving disproportionately through expense reports rather than procurement. Its survey of 218 IT leaders found 78% reported unexpected charges tied to consumption-based or AI pricing in the prior 12 months, and 61% had cut projects because of unplanned SaaS cost increases.

Two practical consequences. First, consumption pricing means a license count is no longer a cost estimate — the same seat can cost very different amounts month to month, so the audit has to capture usage-based spend separately rather than annualizing one invoice. Second, consolidating AI tools is not the same kind of move as consolidating ordinary SaaS: prompts tuned to one model, fine-tunes that don't export, and embeddings tied to one vector space mean the exit cost is higher than the subscription suggests. We cover that layer separately in what actually breaks when you switch AI vendors.

Bottom line

Sprawl isn't primarily a tidiness problem; it's a measurement problem that corrupts the inputs to every switching decision downstream. If roughly a third of licenses are unused and most spend sits outside IT's direct view, then the "current cost" in a tool comparison is close to a guess — and usually a high one. Audit the stack, rightsize what's real, and run the switching calculation against the trimmed baseline. Some proposed switches survive that; the ones that do are the ones worth the migration.