Dawson Whitfield had been designing logos for roughly twelve years and had made about 150 of them for clients when the process started to feel absurd to him. Building 30 mockups took ages, the back and forth with a small business owner dragged on for weeks, and the logo that came out the other end was often simple enough that the whole exercise felt wasteful. He was a designer who could also code, and he had spent time as a UX designer at Weebly working on the website creation experience. So instead of complaining about the process, he tried to replace it.

The only piece of validation he did before building was a Google Trends lookup. It told him that people searched for "logo maker" about three times more often than "website builder". He kept that tab open and, by his own count, went back to it around 200 times whenever motivation dipped. There were no interviews, no landing page test, no waitlist. The second signal he trusted was that the existing logo makers were, in his view, poor enough that a much better product had obvious room.

The dates in this story come from Whitfield himself. Answering a question under his Indie Hackers interview, he wrote that he started working on it on September 1, 2016, launched a beta on November 15, and posted on Product Hunt on November 22. That is two and a half months from first line to public launch, done alone, alongside a little freelancing for existing clients that helped pay for it along with his savings. All day counts here are measured from September 1, 2016.

The build was less exotic than the label suggested. The stack was PHP, MySQL and jQuery. Logos were assembled out of ingredients: fonts, colors, layouts and symbols. The machine learning part tracked about 80 types of user action, from which inspiration someone picked to which logos they favorited, which fonts they swapped and what they eventually bought. A daily job weighted those actions and turned the heavy ones into rules about which ingredients belong together. Whitfield was blunt in the Hacker News thread that the algorithm was still in its infancy and mostly worked as an idea source. What carried the product was the second half: an editor that showed variations as real logos rather than dropdown menus, and mockups of the logo on business cards, signage and shirts.

The pricing model did the rest. Anyone could design as many logos as they wanted for free, and paid only to download: $20 for a basic package or, more commonly, $65 for the premium package with vector files. Whitfield's own explanation for why people paid $65 for something as simple as styled text was presentation. People saw the finished thing before spending anything.

Product Hunt on a Tuesday, posted by someone with reach, was the entire launch plan. Logojoy made $450 on launch day. A week later, on November 29, Courtland Allen published the Indie Hackers interview and Hacker News picked it up, sending enough traffic that Whitfield spent the day apologizing in the thread for logos loading slowly. The headline figure at that point, $15,000 a month, was an extrapolation from launch week sales, and Whitfield said in the comments that it was calculated without the Hacker News spike. That is worth flagging: the $15,000 number was a run rate, not a month that had happened.

What happened next was not a spike that faded. Traffic after launch was almost entirely word of mouth, and it spread further than he expected. Within a day, Japan was the biggest market at about 25 percent of all traffic. By early 2017 the site was seeing roughly 5,000 visitors a day, about a quarter of whom finished a logo and signed up, and about 0.5 percent of signups bought. He had started buying AdWords clicks at around $0.20, spending about $250 a day for roughly 1,800 extra visitors. In February 2017, Logojoy made nearly $70,000 in the month, a run rate near $840,000 a year, about six months after he started building.

The pricing kept generating growth after the milestone. Writing later about how Logojoy priced itself, Whitfield described pulling the social media kit out of the premium package and selling it as a $9 add-on instead, which added about $30,000 a month. Letting premium customers change their logo for 72 hours after purchase, a response to the fact that more than half of inbound support contacts were change requests, lifted conversions by almost 15 percent. More than 20,000 logos sold in the first year.

The part he said he would change was the beginning. Asked what he would do differently, his answer was to get beta users earlier, because watching people use the site was what made him finish features and fix bugs.

The later chapters are less comfortable and worth knowing. Logojoy renamed itself Looka in April 2019 and widened its offer to full brand packages. According to reporting by BetaKit, the domain switch cost about 80 percent of organic traffic instead of the 20 to 30 percent the team had planned for, cut revenue in half, and led to two rounds of layoffs that October. Whitfield's conclusion afterwards was to simplify and go back to the one product that worked, the logo maker.