Pinboard: 11 years of revenue in one table, a $259K peak and three straight years of decline

Pinboard took in $212K in 2020, down for the third year running from its 2017 peak of $259K. Running the site costs roughly $17K a year by the founder’s own account, with zero outside funding and zero employees. Converted to yen in the frontmatter, that is 1.8 times the median among the 19 web services on this site that disclose monthly revenue, and about half the median of our 63 United States cases. The business is usually introduced as a profitable subscription service kept alive by one person for eleven years, yet the table it publishes does not slope upward. Bookmarks climbed from 3.5 million to 244 million while active users alone fell from 29,000 in 2017 to 19,000 in 2019, two thirds of the earlier number. The cause founder Maciej Ceglowski names is neither a competitor nor a technical failure. It is his own absence from the business for two years.

The same table, every July 9

Pinboard marks its July 9, 2009 launch date by posting nearly the same set of statistics each year. Revenue, active users and bookmarks from the table in the July 2020 post:

The same table carries a funding row where all eleven years read zero, and an acquisitions row that shows a single 1 in 2017. The 2018 statistics are missing from that year because the anniversary post was forgotten, and they were filled in a year later.

The economics of a site that costs $17K a year

Revenue first appeared in the seventh anniversary post in 2016, and the reasoning came with it: he had stopped being afraid of competitors, and he wanted to encourage people considering a one-person or zero-person business. The cost line sits in the same paragraph. “The site costs something like $17K/year to run, so you can make a good living at this artisanal SaaS stuff.” Against $234K of revenue, fixed costs of $17K leave the rest as his own pay and taxes.

Pricing is traceable too. The June 2017 post quotes $11 a year. The current About page lists $22 a year for the basic plan and $39 a year for the plan that adds full-text archiving and search, with discounts for multi-year terms. Moving from a one-time fee to annual subscriptions in 2015 pulled revenue down from $193K to $160K, and in 2016 the first wave of renewals brought it back to $234K. That dip is not a failure, it is the trace of a change in accounting shape. Any business that shifts from one-time payments to subscriptions digs a one-year hole on the way across. Pinboard’s table is an unusual record of that hole filling back in within twelve months.

Bookmarks kept growing, people did not

Between 2017 and 2019 stored bookmarks went from 173 million to 244 million and archived data from 31.8TB to 82TB. Active users alone slid from 29,000 to 19,000. The people who stayed used it harder, newcomers stopped arriving, and existing customers left.

Ceglowski explains the drop in a single line: he spent all of 2017 on tech organizing, all of 2018 raising money for the congressional elections, and customers grew irate. His 2017 post says the same thing from closer range, describing travel that began the previous November and support email left unanswered for weeks at a time.

This is the heart of a one-person business. An income statement showing $259K of revenue against $17K of fixed cost looks like a business with almost no cost structure. What actually shrank was not an expense line but the hours he spent answering support mail. The real cost of a one-person business is not the server bill, it is the time the owner is not there, and that figure never reaches the books. The inventory here is neither disk nor bandwidth. It is the fact that email gets a reply.

Even so, revenue fell only 18% over three years. This next part is our inference rather than his: for customers who had deposited years of bookmarks and archives, the cost of moving out exceeded the cost of staying. An asset that grows harder to leave as it accumulates absorbed the owner’s absence.

An individual bought the site that had been sold five times

In June 2017 Pinboard acquired Delicious, the single 1 in the acquisitions row. Ceglowski had written Pinboard in 2009 partly out of frustration with a Delicious redesign, so what he bought was his long-standing rival. The price was never disclosed.

By his own summary, Delicious was founded in 2003, took investment from Union Square Ventures in 2005, sold to Yahoo later that year for a reported $15M to $30M, was announced for sunset by Yahoo in December 2010, passed to the YouTube founders in 2011, to Science, Inc. in 2014, to Delicious Media in 2016, and to one person in 2017. Sold five times, and the final buyer was a business with no employees. The announcement closed with a warning not to attempt to compete with Pinboard.

The purchase was not a growth move. He wrote that he bought it in part so it would not disappear from the web, which makes it an archivist taking custody of a competitor. Judging by the revenue table, the numbers started falling the year after the deal.

The headwinds now

The January 2025 post shows both that the business is still running and that it faces a kind of load it never had eleven years ago. Bot crawling, most of it originating in China, imitates ordinary browser behavior and spreads across tens of thousands of IP addresses, so throttling by address does nothing. User and tag pages are expensive to generate, so to keep the whole site from going down he moved public pages behind a login. That contradicts the site’s own design principle that public things should be public, and users who relied on it to share links were upset. A CAPTCHA has since restored public access, with a note that other measures will follow if the bots outsmart it.

The Twitter archiving feature is also dead, killed by API pricing at what is now X. He is open to a Bluesky equivalent, and has asked users to describe how it should work because he does not use Bluesky himself. Any feature that plugs into an outside service turns that service’s price list into a shutdown switch.

A lot remains unpublished. Profit, churn, the mix between the $22 and $39 plans, the Delicious purchase price and every year of revenue after 2020 are absent from the table. The $17K cost figure dates from 2016, and the archive has since grown to 82TB, so there is no guarantee it still holds. Between the first support email ever received, which angrily demanded a refund, and the line in the tenth anniversary post calling a one-person business an exercise in long-term anxiety management, the entire record of this company exists in the founder’s first person.

How far this carries

Among one-person businesses that run without advertising for a long time, this site covers Photopea, a browser image editor run solo for 13 years on $3M of revenue against $12,600 of expenses, BuiltWith, one person reaching $14M a year over 19 years, and Carrd, an anonymous solo developer at $2M ARR with zero marketing spend. Pinboard’s $212K is the smallest of the group. What stands out among our cases is not the scale but eleven years of the same table published in the same shape.

• Fixed costs of $17K a year stay profitable only while users treat a service that adds no features and hires nobody as a specification rather than a defect
• Surviving the revenue dip from switching to annual subscriptions works only when cash lasts until the renewal wave arrives and usage habits never break in between
• Revenue falling just 18% through two absent years happens only when the volume of data customers have deposited is itself the cost of leaving

The limits belong here too. Pinboard launched into the one-time tailwind of the December 2010 Delicious shutdown, and the growth from 2,800 users in July 2010 to 16,000 in July 2011 cannot be manufactured after the fact. People from the earliest days of Delicious gave him both pricing advice and a moral blessing, which he credits in his anniversary posts. Every figure in the table is self-reported with no third-party audit. And because updates stopped in 2020, whether the decline in active users has bottomed out is not something the public record can answer.