From selling croissants on the train to generative AI
We asked Dorian Selz, Co-Founder and CEO at Squirro (a current BuildGroup portfolio company), to share his fascinating serial-entrepreneur journey with us. Dorian shared his story from selling beer and croissants on trains in Bern, to studying econometrics and founding multiple companies (including a note-taking app that competed with Evernote).
We explored Dorian’s journey and the founding of Squirro—an AI-driven platform helping companies turn unstructured data into actionable insights. Squirro’s client list includes powerful global brands in highly regulated sectors: Bank of England, Henkel, Bertelsmann, Sainsbury’s, European Central Bank, and many more. Without further ado, our conversation with Dorian.
My grandparents on my mother’s side ran a precision mechanical company that became the first-ever gold supplier to IBM in Europe in the 1960s. Back then, computers weren’t just made of chips. My grandparents played a role in that era of innovation, manufacturing fine mechanical elements that were integral to the mainframes.
My two brothers and I visited their workshop quite often, and I was always inspired by a certificate that hung in my grandfather’s office. He had grown up in a family of eight siblings and founded the company before World War II – a testament to his resilience and entrepreneurial spirit. That legacy left a lasting impression on me and shaped my own journey as an entrepreneur.
Swiss trains are spectacular! Back then, they had small carts going up and down the aisles selling coffee, croissants, sandwiches – whatever passengers might need. This became my first taste of entrepreneurship, a way to earn my own money and be my own boss.
The business was brutal. In the morning, I’d receive a fully stocked trolley, and by the evening, whatever remained had to be returned. My earnings came solely from commission – there was no fixed salary. The difference between what I sold and what I owed determined my take-home pay.
Sunday evenings, when the military returned to the barracks, we would sell beer like crazy. Those routes were the most profitable.
We actually had a stack-ranking system to track our performance, comparing our sales against both the daily and monthly averages for each route. It was an extremely competitive environment, but for a so-called non-educated job, it was fantastic. We earned about fifty to sixty bucks an hour in today’s money – a record at that time.
To a large extent, I still sell croissants.
I went to the University of Geneva to study econometrics but I quickly realized that spending my days producing economic research papers to be read by no one but myself wasn’t for me.
Then, in the mid-90s, I received an offer to pursue a PhD at the University of St. Gallen, a reputable business school on the other side of Switzerland. I completed my PhD at the Institute for Information Science, where we were literally wiring the first web pages for companies like UBS.
From there, we launched a spin-off called Namics, which grew into the largest business transformation agency in Switzerland and later expanded into Germany. What started with three founders soon became a seven-person partnership, and together, we scaled the company to 800 people, making it the largest agency of its kind in the region.
Toward the later stages of Namics, one of my clients and I co-founded our next company, which digitized all of the Yellow Pages in Switzerland. Back then, there were still telephone books, and Google Maps didn’t exist. But anyone paying attention could see that Google wouldn’t ignore the Yellow Pages market for long – and for good reason: it was incredibly profitable, with margins of 45–50%. The only industries with higher profit margins were... well, let’s just say, not exactly legal.
Jokes aside, the business model was straightforward. It was built on small transactions that, when scaled, amounted to a massive business. For many local businesses, like hair salons and restaurants, Yellow Pages were the only viable way to advertise their services within their region. Local newspapers were often too expensive, leaving Yellow Pages as the best and sometimes only option. A small business might pay $1,000 or $2,000 for an ad, but with around 130,000 customers, that added up quickly.
localsearch was eventually acquired by Swisscom, a major telecom and media company in Switzerland that had always been a co-owner of the Yellow Pages business. As they took over the majority stake, their CEO recognized that digital transformation would inevitably overtake the traditional print business. Within just four years, we became the largest website in Switzerland—even surpassing Google locally at the time.
Unfortunately the CEO passed away, and after that, my two co-founders and I decided to move on. We took a few key lessons from that experience and launched our next venture, Memonic, a note-taking application. However, we ultimately lost flat-out to Evernote.
At the time, we had a nearly identical value proposition and were even a few months ahead of them. There was a lot of hype around the new freemium business model, which turned out to be far less sustainable than many believed. Evernote’s free-to-paid conversion rate was around 4%, while ours was slightly better at 8%. In practice, that still meant that for every 100 users, 92 of them were essentially getting a free lunch. In the long run, that just wasn’t a viable business.
At Memonic, we were essentially building a note-taking application that could talk back to you with your own data. To make that possible, we had to truly understand what was in the data. That meant dealing with vast amounts of unstructured information and developing the ability to differentiate between different types of content.
In that process, we had already started working on what is now widely known as vectorization, the fundamental concept that underpins large language models today. At its core, vectorization is about transforming words into mathematical representations, which is exactly what modern AI does.
This approach wasn’t new. It dates back to a mathematician couple, Robertson and Spärck Jones, from the 1960s who laid the foundation. We were simply building on their work. And when Memonic went belly up, we didn’t stop. We kept pushing forward—and that ultimately led to the creation of Squirro.
It wasn't driven by a personal challenge, but rather a key insight: as more devices emerge and more content appears online, the volume of data keeps expanding. The real challenge is making sense of it all, and being able to differentiate between unstructured datasets in meaningful ways.
That's why we kept going. Instead of returning to corporate life after Memonic’s failure, we saw an opportunity to build the technology needed to solve this problem. And that’s what led to Squirro.
By working with vastly different companies, some of whom operate in highly regulated environments, with all sorts of data sets and business needs, we were able to figure out how to deliver the safest and most performant intelligence tool at scale, permission-enabled and tailor-made for these complex enterprise situations.
Three of my four companies have had the same co-founders: Toni and Patrice. We go back more than 25 years now. Patrice was actually my apprentice in the late ‘90s – he was just 16 or 17 at the time. And Toni was originally a partner at a company we worked with. He later joined me, and from there, the three of us built all our companies together.
What’s interesting about my relationship with Tony and Patrice is that, professionally, we’re joined at the hip, but we're not necessarily the best of friends. We each have our own families, our own friendships outside of work, and I think that’s actually the secret to our longevity as co-founders. Over the years, we’ve each gone through different life stages. Patrice got married, I got married, we had kids, Tony had kids. Through it all, we’ve remained close but maintained a healthy distance, and that balance has helped us navigate all the ups and downs that inevitably come with building multiple companies together.
There are many easier ways to earn a decent living than to do what we do as founders. You're constantly forced to do things that, from an outside perspective are actually bad choices, but from the inside perspective are the right choices. If you look at it [entrepreneurship] from an outcome perspective, or even from a rational perspective, any entrepreneur is worse than a lottery player. A lottery player can keep buying tickets to increase his or her chances of making it big, even though the chances of success are minimal. An entrepreneur like me, who is bound to the company, is bound to one ticket and statistically, the chance of success is minimal. You need to be crazy to do this.
This is where I need to compliment Eric and Lanham at BuildGroup–it takes guts and bravado to invest and partner with a Swiss company with uncertain perspectives out of Texas. Their thesis at the time (in 2021) was that the next frontier for AI and applications that leverage AI was shifting from structured to unstructured data.
The information in this blog post is provided in good faith without any warranty. It does not constitute investment advice, recommendation, or an offer of any services or products of BuildGroup Management and it is not intended to provide a sufficient basis on which to make an investment decision. This document is provided for educational purposes only. Discussions of current or former BuildGroup portfolio companies are intended for educational and discussion purposes only. Any portfolio company discussed has been selected based on objective, non-performance based criteria.
This content does not constitute or form part of an offer of any investment advisory services of BuildGroup Management, LLC, nor does it constitute or form part of an offer to issue or sell, or of a solicitation of an offer to subscribe or buy, any securities or other financial instruments, nor does it constitute a financial promotion, investment advice or an inducement or incitement to participate in any product, offering or investment.
©BuildGroup 2024
©BuildGroup 2024
