The PMC audit starts with your software stack. Not your ads, not your SEO. The stack, because it is the one place where recoverable budget sits in plain sight, and cutting it costs you no capability and no headcount.
What follows is the published research on software waste, with one caveat stated up front rather than buried: almost all of it is published by companies that sell software management tools. That does not make it wrong. It does mean the numbers deserve to be read carefully, and it turns out that some of them do not survive being read carefully, including by the standards of the firms publishing them.
Start with the number that is actually solid
Zylo has more than 40 million SaaS licenses and $75 billion in spend under management, which makes it the largest published dataset on this question. Its 2026 SaaS Management Index puts the average organisation's annual waste on unused licenses at $19.8 million.
That is the figure I would take to a board. It appears consistently across Zylo's own publications, it is explicitly attributed to a named index year, and the underlying dataset is enormous.
The number that is not solid, and why that matters
The percentage of licenses that go unused is quoted everywhere, including in the earlier version of this article, and it does not hold up. Zylo's own pages currently give different answers: the 2026 index release states 36 percent, while Zylo's license waste solutions page and its own explainer blog both state 46 percent, neither of them tied to a stated index year. Earlier indexes said 51 percent and 53 percent.
I am not going to pick one and present it as fact. The honest position is that Zylo's published unused-license percentage has moved by seventeen points across four years and currently disagrees with itself by ten points across two pages on the same website.
If the vendor with forty million licenses under management cannot state one figure consistently, treat every SaaS waste percentage you read as directional. The dollar figures are better sourced than the percentages, and your own invoice is better sourced than either.
This is not a reason to ignore the research. It is a reason to use it for direction and to use your own subscription list for decisions.
What is defensible about the shape of the problem
Strip out the contested percentages and a consistent picture remains across three independent datasets.
App counts are rising again. BetterCloud's State of SaaS put the average company at 106 applications in 2025, down from 112 in 2023. The 2026 report reverses that: 118 applications, up 11 percent year over year, with the mid-market moving from 116 to 164. The consolidation wave that trade press announced two years ago did not hold.
Nobody central is buying any of it. Per Zylo's 2026 index, business units now control 81 percent of SaaS spend. IT directly manages 15 percent, and individual employees the rest. Purchasing happens without anyone holding a cross-stack view, which is the mechanism that produces overlap in the first place.
Visibility is getting worse, not better. Flexera's State of ITAM reported complete visibility across the technology stack at 47 percent, then 43 percent, and in its 2026 report 36 percent. Over the same period reported SaaS wasted spend rose by ten percentage points. Organisations are losing track of what they own faster than they are cleaning it up.
Spend per head is climbing, and AI is why. Zylo's 2026 median is $9,455 per employee. AI-native application spend grew 108 percent year over year overall, and ChatGPT is now the single most expensed application. If your software bill grew this year and nobody can point to a decision that caused it, that is the likely reason.
Where redundancy clusters
Zylo's 2026 figures for duplicate tooling, per organisation:
- Online training and learning management: 14 duplicative tools
- Project management: 10 overlapping tools
- Team collaboration and messaging: 10 parallel tools
The pattern is the same in every one: categories where different departments adopted independently, and nobody checked what already existed.
The part nobody publishes, and you should know it
Here is the honest limitation of everything above. None of this data describes a small business.
Zylo's smallest reported band is 1 to 500 employees, and that band averages $11.5 million in annual spend across 152 applications. Cledara's 2025 Software Spend Report reaches further down, and its smallest band, 0 to 20 employees, still averages $121,336 a year in software. Its 50 to 100 band averages $193,716, and its 100 to 200 band averages $251,119, of which it reports $89,033, or 34 percent, as waste. Above 200 employees Cledara puts waste at 48 percent of software spend.
A five-person salon spending $600 a month is not in any of these datasets. The percentages may generalise downward and I suspect they do, but no published study I can find tests it, and I am not going to multiply someone else's enterprise ratio by a local business's invoice and call it research.
What this looks like in specific verticals
What does exist below the research floor is the pattern that shows up in every stack audit. Prices below are the vendors' own published rates at the time of writing, which is worth stating because most articles on this subject quote prices nobody has checked in two years.
Restaurants and hospitality
A typical restaurant runs a POS, a separate online ordering platform, a reservations tool, and two or three delivery integrations each with their own dashboard. The specific overlap worth checking: Toast now ships Toast Tables, which covers ground an existing OpenTable subscription already covers.
OpenTable's entry tier is $149 a month plus $1.50 per network cover after the first thirty days, with Core at $299 and Pro at $499. Toast does not publish a price for Tables or state anywhere on its own site whether it is bundled, so the only way to know what you are paying for it is to ask your rep and get the answer in writing. Many operators pay both because nobody checked when the feature appeared.
Salons, barbershops, and fitness studios
The same three products turn up constantly, often two of them in parallel:
- Mindbody publishes "starting at $79 per month per location." Its three plans are quote-based, so the real number is whatever you negotiated.
- Vagaro publishes about $24 a month for one location with one bookable calendar, with each additional employee calendar adding $10 a month up to seven, after which further calendars are included.
- Square Appointments has a genuine free tier at $0 per location, then $49 for Plus and $149 for Premium.
All three hold a fragment of the client list, and the fragments do not reconcile. That is the real cost, and it is larger than the subscription line.
The four checks
Every active subscription runs through these, in this order:
- Duplication. Does this tool do something another tool you already pay for also does? If yes, pick one. The rule is not to keep the cheaper one. It is to keep the one your team actually opens.
- Data visibility. Are the numbers this tool produces ones you can act on, or are they locked inside its own dashboard? If the data never reaches your CRM or your reporting, you are paying for figures you cannot connect to revenue.
- Customer record. Can a customer interact with your business through this tool without that interaction landing on a contact record you own? If yes, the tool is building someone else's database at your expense.
- True cost. Monthly fee plus the time spent training, managing and reconciling it against everything else. A $30 tool with four hours of monthly admin attached is not a $30 tool.
Common questions
Is the Zylo data from enterprise companies only?
Effectively, yes. Its smallest band tops out at 500 employees and still averages $11.5 million in spend across 152 applications, which is not a small business by any definition that matters to a local operator. Cledara reaches down to 20 employees. Below that, nobody is publishing.
Our team is small. Does the overlap tax apply to us?
The mechanism does, and it is arguably worse: at small companies nobody has a cross-team view at all, so every manager buys what they need and no one ever revisits it. What I will not do is quote you a percentage, because the studies that produce percentages do not cover businesses your size.
What is the first tool to cut?
The one duplicating a feature already included in something you pay for. Most common: a standalone scheduling tool running alongside a POS or CRM that already schedules. Second most common: a standalone email tool alongside a CRM that already sends email.
How do I find what we are actually paying for?
Pull the last three months of card and bank statements and filter for recurring charges. List every one. Then, for each, ask whether another tool does the same job, and who the primary user is. If you cannot name a person, cancel it at the next renewal. This exercise takes an afternoon and needs no research report at all.
