The average company now spends about $10,800 per employee per year on software (Bright Interaction, 2026).
Not on computers. Not on salaries. On subscriptions.
That figure was $8,500 in 2023. It is climbing fast.
Now the part that should stop you.
Somewhere between a quarter and half of that spend is wasted.
The sources disagree on exactly how much, and that disagreement is itself informative. Here is the full picture. 💸
🎁 Test Before You Buy — Free Trials by Category →
🧾 Principaux résultats en bref
| Mesure | Chiffre | Source |
|---|---|---|
| SaaS spend per employee, 2026 | About $10,800/year | Bright Interaction (2026) |
| Same figure in 2025 | $9,643 | Bright Interaction (2026) |
| Same figure in 2023 | $8,500 | Bright Interaction (2026) |
| Small business total spend | $10,000–$50,000 | CloudNuro (2026) |
| Licences unused or underused | 25–30% | Zylo (2026) |
| Alternative estimate | Only 54% of licences used | Ramp (2026) |
| Global annual waste | About $45 billion | Zylo (2026) |
| Recoverable through cleanup | 23–30% of spend | Zylo (2026) |
📈 The Climb Nobody Budgeted For
Per-employee software spend rose from $8,500 to $10,800 in three years.
That is roughly 27% growth while general inflation ran far lower.
📊 Annual SaaS spend per employee
Why it rises without anyone deciding
Almost nobody sits down and doubles the software budget.
It happens through small, individually reasonable choices.
A team adds one tool. A vendor raises prices 8%. A plan gets upgraded for one feature.
Nobody approves the total, because nobody sees the total.
The per-employee framing matters
Total spend rising can be explained away by hiring.
Per-employee spend rising cannot.
It means each person costs more to equip than they did last year.
What this means for a small business
Small businesses typically spend $10,000 to $50,000 in total (CloudNuro, 2026).
At the top of that range, software is a serious line item.
It is often larger than rent for a small remote team.
🗑️ The Waste Problem, Honestly Measured
Here the sources genuinely disagree, and the spread is wide.
| Source | Waste estimate | What it measures |
|---|---|---|
| Zylo (2026) | 25–30% of licences | Unused or significantly underused |
| Ramp (2026) | 46% wasted | Only 54% of licences used |
| Breeze (2026) | 56% of licences unused | Broadest definition |
Those cannot all be describing the same thing.
Why the estimates differ so much
It comes down to what “unused” means.
Never logged in? Logged in once six months ago? Uses 10% of features?
Each definition produces a very different number.
The strictest definition gives 25%. The loosest gives 56%.
Which figure to use
For planning, use the conservative one.
If a quarter of your software spend is genuinely dead, that is already a large number.
Assuming half may be true, but it is harder to defend in a budget conversation.
🥧 Where software money goes, using the conservative estimate
🔍 Where the Waste Actually Comes From
Waste is not random. It clusters in four predictable places.
| Cause | Why it happens | How hard to fix |
|---|---|---|
| Licences for people who left | Offboarding misses software | Facile |
| Duplicate tools | Two teams solved the same problem | Moyen |
| Overprovisioned tiers | Bought headroom never used | Facile |
| Role changes | Person moved, licence did not | Facile |
| Trials that converted silently | Nobody cancelled | Easy, once found |
Three of those five are easy wins.
The single biggest is licences belonging to people who no longer work there.
The offboarding gap
When someone leaves, IT disables their email and laptop.
Nobody thinks about the twelve subscriptions in their name.
Those keep billing, sometimes for years.
The duplicate-tool problem
Marketing buys one design tool. Product buys another.
Neither knows about the other, because neither purchase was large enough to need approval.
Small purchases avoid scrutiny, which is exactly why they accumulate.
Overprovisioning
Plans are often bought one tier above actual need.
The reasoning is sensible: leave room to grow.
But growth often does not come, and the tier never gets reviewed.
🧮 What This Costs a Real Business
Let us put actual numbers on it.
Take a ten-person business at the 2026 average of $10,800 per employee.
That is $108,000 a year on software.
At the conservative 27% waste rate, roughly $29,000 is doing nothing.
📊 Annual waste by team size, at 27%
Put that in context
For a ten-person business, $29,000 is a meaningful sum.
It is a junior salary. It is a year of marketing budget.
And recovering it requires no new customers and no extra sales.
Compare it to the alternatives
To generate $29,000 of extra profit through sales, you need far more revenue.
At a 20% margin, that means $145,000 of new business.
Cutting waste is the same result without finding a single customer.
Why nobody chases it
Because it is invisible and boring.
Nobody gets promoted for cancelling subscriptions.
The saving does not show up as revenue, so it attracts no attention.
📋 The Audit That Takes One Afternoon
You do not need software to find software waste. You need a card statement.
| Étape | Ce qu'il faut faire | Time |
|---|---|---|
| 1 | Export 12 months of card and PayPal statements | 15 min |
| 2 | Highlight every recurring charge | 30 min |
| 3 | List them with cost and owner | 30 min |
| 4 | Mark each: essential, useful, unknown | 20 min |
| 5 | Cancel everything marked unknown | 45 minutes |
| 6 | Diarise a repeat for six months | 2 minutes |
Under three hours, and step four is where the surprises live.
Most people find at least two charges they cannot identify at all.
The “unknown” category is the point
If you cannot immediately say what a tool does and who uses it, that is your answer.
Cancel it. If someone needs it, they will say so within a week.
That sounds reckless. In practice, almost nobody complains.
Check the app store subscriptions too
Card statements miss one common hiding place.
Subscriptions bought through phone app stores bill through the store, not the vendor.
They appear as a single lump from Apple or Google.
Open your phone’s subscription settings and read that list separately.
Most people find something there they forgot entirely.
The annual-billing trap
Annual subscriptions do not appear on a monthly statement.
They hide in one month of the year.
That is why the audit must cover twelve months, not three.
Do it before renewal season
Most annual renewals cluster in January and at financial year end.
Audit six weeks before, when you can still cancel.
Auditing the day after a renewal costs you a full year.
🧾 What Categories Eat the Budget
Spend is not spread evenly. A few categories dominate most software bills.
| Catégorie | Typical share | Waste risk |
|---|---|---|
| Communication and collaboration | Largest single block | Low — used daily |
| Design and creative | Significant | Medium — seat-based |
| Marketing and analytics | Significant | High — overlapping tools |
| Security and backup | Modéré | Low — but check tiers |
| Finance and admin | Modéré | Faible |
| Specialist or one-off tools | Small individually | Highest — forgotten fastest |
The bottom row is where most waste hides.
Individually small charges never trigger a review, so they survive indefinitely.
The marketing stack problem
Marketing accumulates overlapping tools faster than any other function.
Analytics, email, social scheduling, landing pages, forms, SEO.
Several of those overlap by 60% or more in features.
Consolidating two into one often loses nothing you actually used.
Seat-based tools deserve special attention
Anything charged per person scales silently with headcount.
Add three people and three subscriptions grow, not one.
Check seat counts against your actual staff list twice a year.
📉 The Price Increase Nobody Tracks
Vendors raise prices, and almost nobody notices.
A tool at $29 quietly becomes $34, then $39 over two years.
Each rise arrives in an email that looks like an update notice.
Per-employee spend rose 27% in three years partly for this reason.
How to catch it
Record what you pay per tool, once a year, in the same spreadsheet.
You are not looking at the amount. You are looking at the change.
A tool that rose 30% while your usage stayed flat deserves a conversation.
The renewal negotiation nobody attempts
Annual renewals are more negotiable than most people assume.
Vendors would rather discount than lose a customer at renewal.
Asking costs one email and occasionally saves 10 to 20%.
🛒 How Waste Gets Created in the First Place
Prevention is cheaper than cleanup. Three habits cause most of it.
Buying on enthusiasm rather than need. A demo impressed someone.
Buying to solve a future problem. The problem never arrived.
Buying without naming an owner. Nobody is responsible for reviewing it.
The named-owner rule
Every subscription should have one person’s name against it.
That person answers one question every six months: still needed?
Tools without owners are the ones that survive forever unused.
The trial-first rule
This is the habit that prevents the most waste.
Never buy software you have not used on real work.
Free trials exist precisely for this, and most people underuse them.
Notre analysis of trial conversion data covers how to run one properly.
The replacement rule
When adding a tool, ask what it replaces.
If the answer is nothing, you are adding cost rather than shifting it.
Additions compound. Replacements do not.
🏢 Why Bigger Teams Waste Proportionally More
Waste scales worse than linearly, and the reason is structural.
| Team size | Main waste driver |
|---|---|
| 1–5 people | Forgotten trials |
| 5–20 people | Duplicate tools across roles |
| 20–100 people | Offboarding gaps |
| 100+ | Nobody has the full list |
Some 74% of IT teams struggle to manage apps and licences (Breeze, 2026).
That is not incompetence. It is a visibility problem.
The list problem
In a small business, one person can hold the whole list in their head.
Past about twenty people, nobody can.
At that point you need an actual document, not memory.
Why a spreadsheet is usually enough
Dedicated software-management tools exist and cost money.
For most small businesses that is solving a $30,000 problem with a $5,000 tool.
A shared spreadsheet reviewed twice a year captures most of the value.
🧑💼 Who Should Own the Software Budget
In most small businesses, nobody owns it. That is the root problem.
Purchases happen across teams, on different cards, at different times.
| Team size | Who should own it | Review frequency |
|---|---|---|
| 1–5 | The founder | Twice a year |
| 5–20 | One named person, part-time | Quarterly |
| 20–100 | Operations or finance | Quarterly |
| 100+ | Dedicated IT or procurement | Mensuel |
The second row is where most businesses get stuck.
Too big for the founder to track, too small to hire anyone for it.
The part-time owner model
Give one person two hours a quarter and the authority to cancel.
Authority matters more than time here.
Someone who can only recommend cancellations will not get them made.
One card, one place
The single most effective control is boring.
Put all software on one payment method.
You cannot audit what is spread across five personal cards and two PayPal accounts.
Why personal cards cause the worst problems
When someone expenses software personally, it leaves no central record.
If they leave, the subscription often keeps billing them, or lapses without warning.
Both outcomes are bad, and both are common.
💡 What to Do With the Money You Recover
Cleanup typically recovers 23–30% of spend (Zylo, 2026).
On a $108,000 budget that is roughly $25,000 to $32,000.
The question is what happens to it.
The trap of instant respending
Recovered budget tends to get spent on new tools immediately.
Within a year, the same waste rebuilds.
Treat recovered money as saved, not as available.
A better use
Consolidate onto fewer, better tools rather than more of them.
Fewer tools means less training, fewer integrations and less to review.
Notre Analyse comparative des offres à vie et des abonnements covers when one-off purchases beat recurring fees.
⚖️ When Keeping an Unused Tool Is Correct
Not every unused licence should be cancelled, and blanket cleanup causes its own damage.
| Situation | Cancel? |
|---|---|
| Nobody has logged in for a year | Oui |
| Used heavily but only in one month | No — seasonal |
| Insurance against a rare event | No — that is the point |
| Holds data you still need | Export first, then cancel |
| Locked-in annual rate you would lose | Check the renewal price first |
| Duplicate of a better tool | Oui |
Rows two and three are the ones cleanup drives miss.
A tax tool used once a year looks like waste for eleven months.
The backup and security exception
Security software often looks unused because it works silently.
Low login counts do not mean low value here.
Judge these on whether the protection is still needed, not on usage logs.
Export before you cancel, always
This is the mistake that turns a saving into a loss.
Cancelling often deletes your data after a short grace period.
Fifteen minutes of exporting protects against months of rework.
The grandfathered-price trap
Old subscriptions sometimes sit on pricing no longer offered.
Cancelling means you cannot return at that rate.
Check what re-subscribing would cost before dropping a legacy plan.
🔬 How Reliable Are These Figures?
This category has weaker data than most, and it is worth saying so.
| Problème | Effect |
|---|---|
| Published mostly by cost-management vendors | Incentive to show large waste |
| “Unused” is defined inconsistently | 25% to 56% spread |
| Samples skew to larger firms | Small business data is thinner |
| Self-reported budgets | Often incomplete |
The first row deserves emphasis.
Companies selling licence-optimisation software publish the statistics on licence waste.
That does not make the numbers wrong. It does mean the generous end of the range should be treated carefully.
Ce qui est bien établi
Even the most conservative estimates put waste at a quarter of spend.
Every source agrees per-employee spend is rising.
And every source agrees the main causes are offboarding, duplication and overprovisioning.
Those conclusions hold regardless of which figure you prefer.
🚫 Ce que ces données ne vous disent pas
It does not give your number. Averages hide enormous variation by industry.
It cannot value partial use. A tool used monthly may still be essential.
It ignores switching costs. Cancelling is not always cheaper than keeping.
It skews to companies with IT departments. Very small firms are underrepresented.
Vendor-published research is not neutral. Check who benefits from the finding.
🏁 La version courte
Software now costs about $10,800 per employee per year, up from $8,500 in 2023 (Bright Interaction, 2026).
Between a quarter and half of that is unused, depending on how you define unused.
Globally the waste is estimated near $45 billion a year (Zylo, 2026).
The causes are boring: people who left, duplicate tools, tiers bought too large.
All three are found by reading twelve months of card statements.
That audit takes an afternoon and typically recovers a quarter of the budget.
Then the real discipline begins: trial before you buy, and name an owner for everything.
One closing thought about why this is worth doing.
Recovered software budget is the cheapest money in a business.
It requires no new customers, no price rise and no extra hours.
It is simply money you were already spending, redirected to something you chose.
Very few afternoons pay as well. 💸
❓ Questions fréquemment posées
How much do businesses spend on software per employee?
About $10,800 a year in 2026, up from $9,643 in 2025 and $8,500 in 2023 (Bright Interaction, 2026).
How much software spend is wasted?
Estimates range from 25–30% to 56%, depending on how “unused” is defined. The conservative figure is safer for planning.
Why do the waste estimates vary so much?
Because definitions differ. Never logged in, logged in once, and uses few features are three different measures.
What causes most software waste?
Licences for people who left, duplicate tools bought by different teams, and plans bought a tier too large.
How do I find my own waste?
Export twelve months of card statements, list every recurring charge, and mark each essential, useful or unknown. Cancel the unknowns.
Why twelve months and not three?
Annual subscriptions only appear in one month of the year, so a short window misses them entirely.
How much can I realistically recover?
Around 23–30% of spend, according to Zylo (2026). On a $108,000 budget that is roughly $25,000 to $32,000.
Should I buy a licence-management tool?
Usually not, below about twenty people. A shared spreadsheet reviewed twice a year captures most of the value.
Which category wastes the most?
Specialist one-off tools, because each charge is small enough to escape review. Marketing stacks come second, through overlapping features.
Should I always cancel an unused tool?
No. Seasonal tools, insurance-type software and anything holding data you need are exceptions. Export first, then decide.
Do vendors really raise prices quietly?
Often. Increases arrive looking like routine update emails, which is part of why per-employee spend rose 27% in three years (Bright Interaction, 2026).
Can I negotiate at renewal?
Frequently yes. Vendors prefer discounting to losing a customer, and asking costs one email.
Who should own the software budget?
One named person with authority to cancel, not just to recommend. Below twenty people, two hours a quarter is enough.
How do I stop the waste rebuilding?
Name an owner for every subscription, and never buy software you have not tested on real work.
📚 Références
Bright Interaction. (2026). SaaS cost per employee 2026Consulté le 8 août 2026 sur https://brightinteraction.com/insights/saas-cost-per-employee/
Zylo. (2026). How much is wasted on SaaS spend? Retrieved August 8, 2026, from https://zylo.com/blog/how-much-wasted-on-saas-spend
Ramp. (2026). The hidden cost of unused software licensesConsulté le 8 août 2026 sur https://ramp.com/blog/unused-software-subscriptions
Breeze. (2026). SaaS tool sprawl statisticsConsulté le 8 août 2026 sur https://www.breeze.pm/articles/saas-tool-sprawl-statistics
CloudNuro. (2026). SaaS statistics 2026: Market trends and industry benchmarksConsulté le 8 août 2026 sur https://www.cloudnuro.ai/blog/saas-statistics-2026
Lectures complémentaires sur ce site
Notre free trial conversion analysis covers how to test tools properly before buying. The Répertoire des essais gratuits lists current trials by category, and our Analyse comparative des offres à vie et des abonnements covers when one-off purchases win.
À propos de cette analyse
Waste estimates here range from 25% to 56% because sources define “unused” differently, and that range is shown rather than averaged. Much of this research is published by companies selling cost-management software, which is stated in the text rather than hidden. Figures were checked on August 8, 2026.
-
What Small Businesses Actually Spend on Software - août 25, 2026 -
Tube Magic vs vidIQ: Unlimited Against Credits - août 25, 2026 -
Is Tube Magic Worth $47 a Month? The Break-Even Maths - 24 août 2026