Dwayne Lafleur and the trouble with a good guess

Dwayne Lafleur had a team working on a perfectly reasonable question: which advertisement should a person see? At Cubics, his advertising network for Facebook applications, engineers and data scientists tried to improve the answer. Then an application developer would change the layout and produce a result the advertising machinery could not match. A sophisticated system had been outperformed by the arrangement of a page. The furniture, apparently, had opinions.

Lafleur carried that observation into Ezoic, the company he founded in 2010. His proposition was that small publishers needed a practical way to test their websites, and that readers' behavior should help decide what worked. It was an engineer's response to a familiar business problem: the person with a strong opinion about a design is often the person who approved it.

Over the following years, the experiment became a company backed by institutional investors. It also became an employer that expanded too quickly. The distance between those two outcomes gives Lafleur's story its interest. He built tools to challenge assumptions, then had to acknowledge the damage caused by one of his own.

A business inside Facebook's young economy

Cubics belonged to an earlier chapter of social media, when independent developers were making applications for Facebook and looking for ways to get paid. Lafleur built an advertising network around that community. The opportunity was specific: developers had users and advertising space; small businesses wanted somewhere to advertise.

On December 6, 2007, Adknowledge announced that it had acquired Cubics. The announcement described a network representing more than 500 application developers and already delivering over a billion impressions a month. Cubics was less than a year old. These were figures for the business at the time of the acquisition, an early indication of how quickly the new Facebook application economy could create demand.

The sale placed Lafleur's work inside a larger advertising operation. Yet the idea that followed came from a limit he had noticed in the product. Better targeting could only do so much if the application itself made poor use of the screen. The developer's choices were part of the economics, whether or not the ad network controlled them.

That distinction matters to understanding his next move. He had already built a business serving people who made things on somebody else's platform. The recurring difficulty was resources. An individual developer could have a useful application without having a research department to investigate every decision about its presentation.

First, buy yourself a laboratory

Ezoic began with websites Lafleur owned. Before asking publishers to trust the software, he used his own properties to try it. He later described a long period of internal testing across different sites and screen sizes. The willingness to spend time with the machinery before selling it is one of the more revealing details of his working life.

John Cole became a central partner in that effort. Together, they invested more than $5 million in web properties, creating a portfolio on which to examine how changes affected earnings. The sites supplied an unusually concrete kind of feedback. A redesign might look promising in a meeting. A functioning publishing business could reveal whether it deserved the applause.

A later account of Ezoic's early development put the testing portfolio at more than 400 founder-owned websites. This was a substantial laboratory, with actual readers encountering actual pages. It also gave Lafleur and Cole direct exposure to the concerns of the publishers they wanted to serve. Their experiments had an electricity bill, figuratively and otherwise.

The name supplied a small joke for anyone with a geological vocabulary. Ezoic means electronic age: the “e” of email combined with “zoic,” as in Paleozoic. Advertising software rarely asks its customers to consider deep time. Lafleur managed to put a little of it on the letterhead.

Let the visitor have a vote

In November 2013, Ezoic announced a $5.6 million Series A led by Balderton Capital alongside New Amsterdam Capital and private investors. The funding supported further software development ahead of a planned launch out of private beta in the first quarter of 2014. The patient internal experiment was preparing to become a service for other people.

The early proposition addressed a practical imbalance. Larger publishers could employ people to create layouts, organize experiments and assess results. A niche website might depend on an owner who also wrote the articles and handled the business. Automated testing offered a way to bring some of that investigative capacity within reach.

The idea also required a publisher to accept a slightly uncomfortable possibility: a familiar arrangement might be holding the site back. Content, usability and advertising occupied the same page. Treating them as separate departments did little to help a reader trying to use it. Ezoic made the page itself an object of repeated examination.

By 2015, the company was an inaugural member of Google's Certified Publishing Partner program. It later received Google's Business Innovation Award. In August 2019, Sageview Capital invested $33 million, with partner Jeff Klemens joining the board. The business had attracted money and recognition around a proposition that began with watching small developers work.

A 2016 account of the Google partnership brought the problem back to everyday publishing. Website owners were having to manage mobile support, speed, usability and income at once. Lafleur emphasized that these sites supported people and businesses. About 65% of the traffic Ezoic was seeing was on mobile. A layout that worked at a desk had to make sense on a phone, too. That is an unglamorous detail with real consequences: a publisher can spend days creating something useful, only to lose a reader in the seconds it takes to encounter an awkward page.

The niche cannot do your work for you

Lafleur's public advice often returns to the individual publisher's choices. In a September 2022 discussion about publishing niches, he challenged the habit of treating a subject category as a ready-made business opportunity. Ezoic's subsequent explanation of his argument emphasized the difference between a popular topic and a website capable of serving its readers.

It is an appealingly inconvenient point. A fashionable niche gives an aspiring publisher something to announce. Building a useful site requires the slower work of understanding an audience, choosing material and improving the experience. The category can attract interest without doing any of those jobs. A crowded restaurant district still has disappointing restaurants.

This is where Lafleur's interest in measurement becomes more personal than a product description. His advice asks owners to examine what their particular site does well. It leaves less room for the comforting belief that somebody else has found a universal setting. A useful experiment has to survive contact with the people it is meant to serve.

Then the forecast met the payroll

On July 17, 2023, Lafleur told employees Ezoic would reduce its team by about 28%. The company had increased its workforce by 62% in 2022. Revenue in the first half of that year had been up nearly 60% from the year before, he wrote, but the company had let costs grow too quickly and had failed to anticipate a major advertising downturn.

“This reversal shows how wrong we got it.”

Dwayne Lafleur, July 2023

The letter placed responsibility with its author. Lafleur described over-hiring as his mistake. For an executive whose business was built around better decisions, this was a public acknowledgment of a consequential bad one. Employees were bearing the cost of an expansion approved above them.

The arithmetic is worth sitting with. Starting with an index of 100, a 62% increase produces 162. A subsequent 28% reduction leaves roughly 117. Even a company larger than its earlier self can undergo a severe contraction. Growth and the experience of working through it can tell very different stories.

The episode complicates an otherwise tidy account of successful experimentation. Testing page layouts offers repeated observations and opportunities to adjust. Hiring commitments carry consequences that cannot be tidied away with a new version of a page. Lafleur's acknowledgment remains part of his record alongside the acquisitions, investment rounds and technical work.

Who owns the audience?

Later in 2023, Lafleur joined Ezoic marketing chief Tyler Bishop on The Publisher Lab to discuss artificial intelligence, search and the prospects for digital publishers. He considered changes on both sides of the business: software producing content and chatbots becoming a way for people to find answers. His interest remained with the publisher operating amid decisions made by much larger companies.

At The Newsletter Conference in New York in May 2025, he appeared on a panel with Industry Dive co-founder Sean Griffey, moderated by Dan Oshinsky. His advice connected newsletters with websites: a publisher who knows an audience through email may also have an opportunity to build advertising revenue on the site those readers visit.

The following month, in writing about open video, Lafleur argued for creators running channels on their own domains and having greater control over their audience relationships. His companion writing examined the risks of depending on centralized platforms. The emphasis had broadened from the arrangement of a screen to the ownership of the business behind it.

There is a recognizable connection to Cubics. His early customers needed a way to earn inside Facebook's application economy. His later writing asks how creators can maintain more independence from platforms altogether. Across those chapters, the people he addresses are trying to turn something they make into a business they can keep running.

Lafleur, originally from Alberta and based in Carlsbad, remains Ezoic's founder and CEO. The work has changed as the web has changed. The useful question underneath it has endured: what would let an independent publisher make a better decision? His own record offers both reasons to ask it and a reminder of how much can depend on the answer.