The two ways to raise revenue per person
The best headcount metric there is, the famous company that doubled it by removing people, and why most businesses cannot copy that and should not try.
Revenue divided by headcount is the most useful single number an owner can track, because it is the one figure that cannot be flattered by working harder. It also has a problem: there are two entirely different ways to make it go up, they lead to opposite kinds of company, and the number itself does not tell you which one you did.
Why this number and not the others
Margin is the obvious alternative and it is easy to move for bad reasons. Defer maintenance, stop replacing vans, delay a hire you needed, and margin improves for two years before the bill arrives. Growth is worse, because a company can grow revenue while getting steadily less efficient at producing it, which is how a business ends up busier, larger, and no better off.
Revenue per person resists both. To move it you either sell more without adding people, or you produce the same with fewer. Both of those are real structural changes, and neither happens by trying harder for a quarter.
It is also legible without a finance function. Two figures, both of which a small company already knows, divided. Track it quarterly for three years and you have a picture of whether the business is getting structurally better, which is not something the profit line will reliably tell you.
The company that made it famous, and what it actually did
The most-cited example right now is Bending Spoons, the Italian acquirer that has bought more than fifty software businesses since 2013 and now owns Evernote, WeTransfer, Vimeo, Eventbrite and AOL. Its revenue per employee went from roughly 1.12 million dollars in 2023 to about 2.57 million in 2025. Slightly more than double in two years, at real scale.
That number gets quoted approvingly a great deal, usually as evidence of what a modern operating model can do. It is worth being specific about how it was achieved, because the mechanism matters more than the ratio.
Bending Spoons took on roughly 1,830 full-time people through the AOL, Eventbrite and Vimeo acquisitions, and has said it expects only a few hundred of them to remain by the end of 2026. That is the denominator falling, deliberately and at speed. The model is explicit about it: buy a business with proven demand, move it onto a shared central platform, and run it with a fraction of the people who ran it before.
Doubling revenue per person by removing people is a real strategy with real results. It is not the same strategy as doubling it by making people more productive, and the ratio cannot tell the two apart.
Worth adding one more figure for honesty, since it complicates the admiration. Forbes noted that the roughly 3.3 billion dollars paid for AOL, Vimeo and Eventbrite sat against a pro forma 2025 profit of about 22 million. Whether the model works is a question with a longer answer than the efficiency ratio suggests, and a metric that looks spectacular while the profit line does not is a reminder that no single number is the business.
The two mechanisms, and which one is available to you
The distinction decides everything about how this applies to a given company, so it is worth drawing side by side.
| Fewer people | More per person | |
|---|---|---|
| What moves | The denominator falls | The numerator rises |
| When it is available | When you have bought several companies that each had their own finance, marketing and support | Any single company where skilled hours are being spent on work that should not need a person |
| What it needs | A shared central platform the acquired business can be moved onto | Capture, quoting, dispatch and billing that run without anyone chasing them |
| The limit | Runs out once the duplication is gone, and the human cost is real | Bounded by how much of the week was administrative in the first place |
| Who it suits | Consolidators buying overlapping businesses | Owners in a trade where the person you would cut cannot be rehired |
For most owners the second is the only real option, and not for sentimental reasons. In a trade or field service business the binding constraint is not payroll cost, it is that you cannot find qualified people at all. Cutting technicians in a market with a shortage of technicians is not efficiency, it is capacity destruction, and the revenue falls with the headcount.
So the useful form of this metric for a service business is: how much more can these specific people, who I could not replace, produce if the administrative drag is taken off them. Which turns out to be a lot, because a large share of a skilled person's week in these companies is not skilled work.
Where the numerator actually comes from
Not from anybody working faster. From enquiries that currently go unanswered becoming jobs, from quotes that take four days going out in one, from maintenance intervals nobody tracks being billed on time, and from hours currently spent on paperwork going back into billable work. Same people, same week, more output.
That is a different exercise from a cost programme and it produces a different company at the end of it. One version has fewer people doing the same work. The other has the same people doing better work, and it is the only one that is compatible with a labour market where the person you would cut cannot be rehired.
It is also worth knowing how much is actually on the table, because "take the admin off them" sounds like a marginal gain and is not. The arithmetic is simple enough to do in your head: if a share of the week is administrative and you recover part of it, billable time rises by that recovered share divided by the time that was already billable. Which produces bigger numbers than most owners expect.
| Share of the week that is admin | Recover a quarter of it | Recover half | Recover three quarters |
|---|---|---|---|
| 20% | +6% | +13% | +19% |
| 30% | +11% | +21% | +32% |
| 40% | +17% | +33% | +50% |
Read the middle of that table. A company where three in ten hours go on paperwork, chasing and rekeying, recovering half of it, gains 21 percent revenue per person. Nobody was hired, nobody left, and nobody worked faster. That is the entire gain, and it is larger than almost any pricing change an owner would dare make.
The honest part is the input. Nobody knows their own admin share, and the guess is always low, because the time is not spent in one block anybody notices. It is spent in four-minute pieces: finding a number, re-entering an address, checking whether somebody rang back, writing the same job details into a third place. Which is why the first move is measuring it for two weeks before deciding anything, and why the answer is usually higher than the owner would have said.
How to actually use it
The metric is only useful with a few rules attached, and without them it will mislead you in predictable ways.
- Count everybody, including the owner, contractors and part-timers converted to full-time equivalents, or the number can be improved by relabelling
- Track it quarterly and read it annually, because it is a structural measure and quarterly noise means nothing
- Record which mechanism moved it, numerator or denominator, every time it changes
- Compare it only against your own history, never against another industry, since the absolute level is set by the sector not by your competence
- Watch it alongside revenue in absolute terms, because a shrinking company can post a rising ratio all the way down
- Check it against gross margin, since revenue per person can also be moved by simply selling more expensive things badly
The last two are the traps. A business losing customers while letting staff go can improve this ratio for several years in a row while dying, and the number will look like progress the entire time. Any metric that improves under both success and failure has to be read next to something that only improves under one of them.
There is a mirror of that trap which gets discussed far less and does more damage, because it punishes the right decision rather than flattering the wrong one. A person hired today produces nothing for a while. They have to be trained, they take time from somebody who was already productive, and in a trade they may not be chargeable at full rate for months. So in the quarter you hire, the denominator rises immediately and the numerator does not. The ratio falls, and it falls hardest exactly when a growing company is doing the correct thing.
An owner watching this monthly will feel that as failure and will hesitate before the next hire, which is precisely backwards. The fix is the same one that handles the shrinking case: read it annually, and write down what you did each time it moved. A dip labelled "took on two apprentices in the second quarter" is information. The same dip with nothing written against it gets misread as decline a year later, by you.
Used properly it does one thing very well: it settles arguments about whether an operational change actually worked. Most improvements in a small company are defended with anecdote, because the profit line moved for six other reasons at the same time. Revenue per person over eight quarters is much harder to argue with, and it is the number I would want to see before and after any claim that a system made a business better.
What it means for what you buy
There is a screening use for this as well, and it inverts what you might expect. A target with unusually low revenue per person is not necessarily badly run, and a high figure is not necessarily good news.
Low revenue per person in a company with good customers and solid work usually means the administrative layer is eating the skilled hours. That is the gap you are buying and it is the whole thesis: a genuinely good business whose people spend a third of their week on things that should not require a person. The full version of that argument is in the product cannot be disrupted, the operations can, and the two-axis version of the screen is in good company, bad ops.
A high figure, on the other hand, means somebody already did the work. The previous owner has been paid for it in the price, and you are buying an optimised business at an optimised valuation, which is a perfectly reasonable thing to do and a completely different transaction from the one above. Both can be good deals. Confusing them is how a buyer pays for an improvement that has already happened. What I look for before either is on the buy page.
The short version
- Revenue divided by headcount is the hardest operating number to flatter, because it barely moves for a quarter of extra effort.
- Bending Spoons took it from about 1.12 million dollars per employee in 2023 to roughly 2.57 million in 2025, largely by taking on around 1,830 people through acquisitions and expecting a few hundred to remain.
- That is the denominator falling. Most businesses cannot copy it, because in a labour-short trade the person you would cut cannot be rehired.
- The available version is raising the numerator: unanswered enquiries, slow quotes, untracked maintenance and paperwork hours, all recovered from the same people.
- The size of the prize: if 30 percent of the week is administrative and you recover half of it, revenue per person rises 21 percent. Nobody hired, nobody released, nobody working faster.
- Read it next to absolute revenue. A shrinking company posts a rising ratio all the way down, and a company hiring ahead of growth posts a falling one while doing the right thing.
Questions I get on this
What is a good revenue per employee figure?
How did Bending Spoons get to 2.57 million dollars per employee?
Can a small service business improve revenue per employee without cutting staff?
Bending Spoons figures from its 2026 public filings and reporting around its July 2026 listing, including Forbes on the gap between the 3.3 billion dollars paid for AOL, Vimeo and Eventbrite and pro forma 2025 profit.