If Your Average Customer Is Worth More Than Ever, Look at Who Stopped Paying You
When AI takes your smallest customers, your average customer looks better than ever. But the customers who stopped paying may have been the door your future large customers came through, so moving upmarket can fix this year's revenue while emptying next year's pipeline.

In 2010, a website went online with an offer that sounded almost too small to be a business. If you needed a logo for a new shop, a paragraph translated into Spanish or a cover letter proofread, you could post the job, pick a stranger somewhere in the world and pay them five dollars to do it. The work came back a day or two later. Over the next decade, for millions of small businesses that could never afford an agency, it was the first time a designer, a copywriter and a translator were within reach.
The price was the idea, and it became the name. The site was called Fiverr. Sellers were later allowed to charge whatever they liked, but the business remained built on a very large number of small buyers paying for small jobs, and by the autumn of 2024 it had 3.7 million of them.
Those small jobs are exactly what a chatbot now does in a few seconds. The obvious story is that AI is taking work away from freelancers, and in a narrow sense it is. But Fiverr's own numbers do not look like a business whose customers are simply disappearing. Over the past two years, while the number of buyers fell by more than a third, the buyers who stayed kept spending more, and by the autumn of 2026 the average buyer was spending more than $370 a year, up from under $300 two years earlier. A company that is losing its market is not supposed to end up with its most valuable customers.
The explanation is that AI did not remove the need. Small businesses still need the logo, the translation and the product description. What AI removed is the reason to pay a stranger five dollars for them, because a shop owner can now do the job in a chatbot, often a free one. The customers who stop paying first are the ones with the simplest jobs, so every one who leaves makes the customers who remain look more valuable. The obvious cost is the small revenue they took with them. The larger cost may be what those small jobs used to lead to.
In 2025, Fiverr's average buyer spent more while the number of buyers kept falling
On November 5, 2025, Fiverr reported that its number of buyers had fallen almost 12 percent in a year, to 3.3 million, while the average buyer's yearly spending had risen by almost exactly the same amount, to $330. Total revenue was up 8 percent, but the marketplace, where buyers hire freelancers and where the five-dollar jobs lived, had already shrunk by 2 percent. What carried the total was Fiverr's separate services business, which sells subscriptions, advertising and software tools mainly to the freelancers themselves and to online shop owners, and which had grown by almost 40 percent.
On the call with investors that day, the chief executive, Micha Kaufman, called the rise in spending per buyer "our strongest growth rate since the COVID era". He did not hide what was happening below it. "The jobs that are being displaced are the very, very simplistic types of jobs," he said, and "those types of services have been very small in size anyway." The report was accurate, and it showed a company that was growing while the business that had made it famous was contracting.
The better average needs no hidden cause. A buyer who used Fiverr for a few small jobs a year spent less than the average buyer, so every one who stopped coming raised the average of the buyers who stayed. In a company with 75 small customers spending $100 a year and 25 large ones spending $1,000, a third of the small ones can leave while the large ones spend 10 percent more, and the revenue line will not move at all, even though a quarter of the customers are gone. Fiverr's public reports cannot show how much of its higher average came from small buyers leaving and how much from large buyers spending more, because it did both. That is exactly why a rising average cannot be read as good news until a company knows which of the two produced it.
A quarter of the customers can leave without the revenue line moving, and the average customer looks better than ever
How to read it. One dot for each of a hundred imaginary customers. The small pink dots each spend $100 a year, and the large black dots each spend $1,000. First, a third of the small customers stop buying, and revenue falls 8 percent while the average rises. Then each large customer spends 10 percent more. Revenue is back where it started, a quarter of the customers are gone, and the average customer is worth a third more than before.
AI closed Fiverr's front door in two places
By July 2026, Kaufman was describing "an ongoing compression in high-volume, low-value, transactional work driven by accelerating AI adoption", and transactions in writing and translation had fallen by more than 24 percent. On October 29, 2026, Fiverr reported 2.4 million buyers, a quarter fewer than a year earlier, an average yearly spend of $373, up 13 percent in a year, and a marketplace whose revenue had fallen by 24.7 percent. This time the services business no longer covered the gap, and total revenue fell 15.8 percent.
In the same results, Fiverr named a second pressure. The move from traditional search to AI-driven discovery would cut traffic to its marketplace more sharply than it had expected, through 2027. "Organic clicks are falling across the web," Kaufman told investors, "and acquisition costs are rising for everyone who depends on them." A shop owner who makes the logo in a chatbot has stopped needing a freelancer. A shop owner who still needs a professional but asks an AI assistant where to find one may never reach Fiverr at all, because the search no longer ends on its front page.
Both are the same door closing in two places. One closes because the first small job is no longer worth paying for, and the other because the buyer who still wants it no longer finds the way in. Fiverr can fight the second. It has built its marketplace into AI assistants such as ChatGPT and Claude, the value of work bought by first-time buyers who arrived from AI tools grew about 64 percent from the previous quarter, and Kaufman called these integrations "the new top of the funnel." Better distribution cannot bring back the first, because a buyer who no longer needs the job will not be found by any funnel.
Fiverr lost more than a third of its buyers in two years, and the buyers who stayed spent more every time
How to read it. Each bar counts the twelve months up to that quarter. On the left, the number of people and companies that bought anything on Fiverr. On the right, what each of them spent on average. The two move in opposite directions in every period. In the most recent quarter, the money Fiverr earned from its marketplace was also a quarter lower than a year earlier.
Sources: Fiverr, third quarter 2025; Fiverr, second quarter 2026; Staffing Industry Analysts, third quarter 2026.
The five-dollar job may have been how the large customers arrived
The answer most companies reach for is to move upmarket, and Fiverr is doing it. In November 2025, Kaufman said "we have consistently prioritized moving upmarket", and in July 2026 he described the company as "aggressively executing our push upmarket." Plenty of companies stop serving small customers on purpose and become more profitable, and a record average after that decision is the plan working.
But a five-dollar job may also have been the start of a longer relationship. A small business owner who paid a stranger for a logo might come back six months later for a website, and a year after that for a marketing campaign, and eventually become exactly the kind of customer worth thousands of dollars a year. Fiverr's public numbers cannot show how many of its large customers started that way, and some will have arrived through its sales team instead. But where the path exists, AI is taking more than today's small revenue. It is closing part of the route by which tomorrow's large customers used to arrive.
That changes what moving upmarket has to achieve. At the end of September 2026, projects worth $1,000 or more made up only 16 percent of the value of Fiverr's completed projects over the previous twelve months. The large work is growing, but it is still a minority of the business. If small jobs really were the way in, it has to replace two things at once: the revenue from the small jobs that are disappearing, and the large customers some of those small buyers would have become.
There is a harder problem waiting at the top. In the Iceman essay, the Customer route of the Five Cs of disappearance began with the biggest buyers, the ones large enough to build what they needed on their own. AI runs the same route from the bottom, because a free tool reaches the smallest customers first. A company that answers by moving upmarket is walking away from the customers who no longer need it, toward the customers most able to stop needing it next. Large customers are not automatically easy for AI to replace, because they often pay for complexity, accountability and integration with their own systems. If that is what they are buying, the move has somewhere to go. If their work is simply the work AI has not reached yet, the company has not found a strategy. It has found a place to wait.
A board should know where its large customers came from before it moves upmarket
A board can see this early, but only if it looks past the average at the customers it was calculated from, and asks three questions before anyone calls a rising average a success.
The first is why the small customers are leaving. Some stopped needing the work because AI now does it for them. Some still need it but no longer find the company, because their search starts somewhere else. Some were let go on purpose, through price rises or closed products. These are three different problems with three different answers, and a company that cannot tell them apart will spend money winning back customers who no longer need it, or give up on customers who simply could not find it.
The second is whether the customers who stayed are really spending more, or only look that way because the smaller ones left. Follow the customers who made up the bottom quarter of revenue a year ago, count how many still pay and how many new small customers arrived, and take the rise in the average apart in money: how much came from customers spending more, how much from new large customers and how much from small customers who stopped paying.
The third matters most. A board should know where its current large customers first came from. If many of them started with a small purchase, the shrinking bottom of the business today is a shrinking top a few years from now, and a plan to move upmarket has to replace the door as well as the revenue.
Fiverr is still named after the door
Fiverr is still called Fiverr. Small businesses still need logos, translations and product descriptions, and the need that built the company has not gone away. Fewer of them pay a stranger five dollars to meet it, and fewer of them arrive through a search that ends on Fiverr's front page. The record average describes the customers who are still inside. The name describes the door many of them may have come in through. Most companies will not have that history written into their name. They will have a record average on a board slide, and will not know how many of next year's large customers were going to be last year's small ones unless someone asks where today's large customers first came from.
Under every essay in this series is the Iceman Test: eleven questions about your own industry that show which of the Five Cs are already moving.
Is your business the iceman?
Industries rarely die because a better rival wins. They disappear because customers quietly move, and in a study of 30 industries, an industry's own numbers were almost always the last place it showed. Eleven questions, about four minutes. Answer for one thing you are paid for, not the whole company.
For example: fix cars, translate documents, sell accounting software, cater weddings.
Sources
- Fiverr: How Fiverr has changed the way businesses get work done
- Fiverr, November 5, 2025: Third-quarter 2025 results
- Fiverr third-quarter 2025 earnings call transcript, The Motley Fool
- Fiverr, July 29, 2026: Second-quarter 2026 prepared remarks
- Fiverr, October 29, 2026: Third-quarter 2026 results
- Fiverr third-quarter 2026 earnings call transcript, TradingKey
- What the 6 Ds of Exponential Technology Don't Show: Why Industries Facing AI Trust the Wrong Numbers
- Nobody Canceled the Iceman