The $21 million most companies spend on software nobody opens

You would notice a floor of empty desks. You do not notice a floor of empty logins.

According to Zylo's most recent SaaS Management Index, the average company wastes roughly $21 million a year on software licences nobody opens. The figure has grown more than 14 percent year on year. The cause is not exotic: only about 49 percent of provisioned licences are actually used, so most organisations are paying for two seats to get one.

That number tends to produce one of two reactions. Either it sounds impossibly large, or it sounds like somebody else's problem. Both reactions are worth examining, because the mechanism behind the waste is almost identical whether you are running fifty people or five thousand.

First, an honest word about that $21 million

It is an average, and averages hide a lot. Zylo's sample includes very large enterprises, and the same index puts large enterprises at around 660 applications against roughly 152 for small businesses. If you run a fifty-person company, you are not quietly setting fire to $21 million. Anybody who tells you otherwise is selling something.

What travels across company sizes is not the dollar figure. It is the ratio. Roughly half of what you pay for goes unused, and that proportion holds whether the total is $21 million or $210,000. The useful exercise is not comparing yourself to the headline number. It is working out your own.

Here is the quick version. Take your annual software spend, whatever you can see of it. Halve it. That is your working estimate for what you are spending on access nobody uses. It will not be precise. It will be closer than your current number, which for most organisations is no number at all.

Why this happens, and why it is not incompetence

Nobody approves a purchase order for software they intend to ignore. The waste accumulates through entirely reasonable decisions, made one at a time, by people doing their jobs properly.

A team of twelve buys twenty seats because they are planning to grow. They grow to fifteen, then someone leaves, then two people move to another department that uses a different tool. Nobody goes back to reduce the seat count, because reducing the seat count is nobody's actual job. The renewal arrives eleven months later as a line item that looks exactly like it did last year, and it gets approved, because it looks exactly like it did last year.

Multiply that by every team and every tool. The average organisation now processes around 211 software renewals a year. That is roughly one every single business day. At that cadence, genuinely interrogating each contract stops being a discipline problem and becomes an arithmetic impossibility.

The part that is invisible to finance

There is a second layer, and it is harder to see. Lines of business now control about 70 percent of software spend. IT manages around 26 percent. More than a third of applications in the average company were bought outside IT entirely.

That software still gets paid for. It just arrives through expense reports rather than procurement. Zylo's 2026 data found an average of 138 expensed applications hiding in expense reports, with more than half of that spend miscategorised in accounting.

So when a finance team pulls the software line and finds it uncomfortable but manageable, that number is frequently incomplete by a meaningful margin. The spend is real. It is simply filed under something else.

The fear that keeps the meter running

Here is the finding that explains why so little gets cut: 76.5 percent of organisations worry that reducing software spend will damage productivity.

That fear is not irrational. Everyone has been in the meeting where a tool was removed to save money and three months of workflow chaos followed. Once burned, organisations develop a hoarding instinct, keeping everything just in case, even when usage data plainly shows nobody has logged in since March.

The problem is that the instinct is applied without evidence in either direction. Tools get kept out of anxiety rather than usage, which means the genuinely load-bearing ones and the genuinely dead ones receive exactly the same protection.

What to actually do about it

Pull login data before you pull the contract. Every major platform can tell you who has signed in and when. Usage evidence turns an argument about somebody's preferences into a conversation about facts, and facts survive the meeting better.

Look for duplicates before you look for savings. The average company runs around 15 duplicate training applications, 11 project management tools, and 10 collaboration tools. Duplicates are the easiest and least contentious cut available, because nobody has to give up a capability, only a second copy of it.

Check the expense reports, not just the procurement ledger. If more than a third of software arrives outside IT, then the procurement ledger is not the whole picture. It might not even be most of it.

Do this before the renewal cycle, not after. One renewal every business day means the window to act on any given contract is narrow and easy to miss. A quarterly review beats an annual panic.

The real cost is not the licence fee

It is worth ending on the thing the money is a symptom of.

A dead licence costs you its price. An active but redundant tool costs you considerably more, because your data is now split across two places, your team has to decide which one is authoritative, and your reporting quietly disagrees with itself. The subscription is the visible cost. The fragmentation is the expensive one.

Cutting the software nobody opens is a good day's work and it frees up real money. But the money was never the whole problem.

You are not just paying for empty seats. You are paying for the distance between the ones that are full.

This is why we are building EvikNova

Every fix above is a good discipline and all of them are worth doing. They share one limitation, though. They make a fragmented stack cheaper to run. They do not make it less fragmented.

This is why we are building EvikNova: to save growing businesses from paying twice. Once in licences for tools that quietly overlap, and again in the hours your team spends making those tools agree with each other.

Rather than another application to add to the count, EvikNova is a single AI-assisted operating layer that runs the functions most growing companies currently buy separately. Customer calls, bookings, sales follow-up, workflows and compliance live in the same environment, against the same record, instead of across six tools that each hold a portion of the truth.

The licence arithmetic changes when the architecture does. You are not tracking seats across eleven products, diarising eleven renewal dates, and reconciling eleven versions of the same customer. There is one system, one record, and one place where the work actually happens. The duplicate tools stop being a procurement problem to manage, because there is nothing left for them to duplicate.

We should be straight about where we are. EvikNova is pre-launch, and we are onboarding a limited first wave now. It has been built for 43 countries and over 100 languages from the outset rather than as something bolted on once the first market was working.

If roughly half your software spend is currently buying access nobody uses, the more useful question is not which licences to cut this quarter. It is why the work needed eleven separate places to live in the first place.

We are building the answer to that question, and we would like you in the first wave.

Join the global early access waitlist