Q1 2025: 12.5%, the best businesses in the world, will AI replace SaaS/VMS and Topicus

Disclaimer: Nothing on this site, including this page, constitutes investing advice. Please do your own due diligence.

In Q1 2025 we returned 12.5%. In the same period, the Nasdaq 100 and SP500 returned -8.3% and -4.6% respectively. In the previous letter, for Q4 2024, I argued for cash as valuations were high and then-president-elect Trump would likely introduce disruptive policies. Looks like that thesis worked out. However, I really don’t have any competency for short term trading. If I ever make a habit of making more short term calls like that, please kindly berate me. Now, onto boring long term investing!

The Best Businesses in the World

If you, from first principles, were to describe the perfect, high quality business, it would likely include the following descriptors: recurring revenue, forever indispensable to customers, impenetrable moat, high returns on capital, asset light, low customer concentration, low labour/supply risk, low operational complexity, no supply chain risk, regulatory capture, network effects, not cyclical, etc. Many very high quality businesses can claim to have a large percentage of this criteria. However, a large concentration of these businesses exist in software. Many of the reasons are well understood: the incremental cost to deliver value to new customers is negligible, software tends to run on top of proprietary data, software scales infinitely, most industries/verticals require some sort of software so there are many applications, and so on. When I think about sectors with the potential for outsized returns for the amount of risk taken, software is unparalleled.

However, clearly, not all software is equal in durability. Napster, Netscape, Lycos, Yahoo, etc. were all software businesses that saw widespread adoption but failed to stand the test of time. Even Google, today, is an example of a business which may yet see disruption by LLMs. By comparison, there is a great deal of software that was written decades ago in forgotten programming languages running today’s financial institutions and they do not seem to have an end in sight. What’s the difference?

The most important distinction I think that’s worth thinking through is whether the software is business or consumer facing. At face value, it is possible to produce a consumer software business with some degree of user lock in. We all still use the same few operating systems: Windows and MacOS. However, consumer businesses have a critical changing factor that have historically killed multitudes of businesses: user experience. This isn’t necessarily just a software problem. As an example, we can see in an era before the Internet, selling music on CDs was a perfectly legitimate business. User experience implies many things: channel to customer, physical medium, design language, and so on. The Internet as a channel to the customer killed Blockbuster, encyclopedias, traditional stock brokers, much of print media, mail services, travel agencies and so on. The phone as a physical medium killed taxi companies, digital cameras, physical maps, GPS and so on. And yet, it is not as if today’s consumer technology is the end state for user experiences. It is likely that the user experience will forever evolve. For that reason, consumer businesses, especially consumer tech businesses, are businesses where the rules and assumptions are constantly changing. That is, in a way, they are businesses built on sand. For an investor, this requires great vigilance and arguably higher discount/hurdle rates.

(Old Reynolds DMS)

Vertical market software/business software, by comparison, does not experience great changes in user experience. Monitors, keyboards and mice have been used for 50+ years. Most serious work has not been disrupted by mobile like it has in much of consumer software. It seems likely that the model of monitor, keyboard, mouse will be used for some time to come. We will get to AI later. An example of VMS is dealer management software (DMS) which serves the car dealership market. The functions of DMS include order management, inventory, accounting, and so on. At the heart of it is essentially some forms/basic UI, a database, and some business logic. Though there are still areas of improvement, the core problems have been solved with many of the rough edges rounded after decades of development. The largest incumbents, at least in the US, are Reynolds and Reynolds (will refer to it as Reynolds), CDK and Dealertrack. They are old school but have very good, mission critical businesses. To put things into perspective, Reynolds was founded in 1866. Estimates vary, but most analysts seem to put the DMS market at something like a $9B market. That market size is large enough to potentially justify a venture backed startup. In 2016, Tesla CIO Jay Vijayan started Tekion, a “modern” take on DMS. Though they have achieved some degree of traction, customer testimonials indicate nearly a decade after founding, is that the product is still a hot mess that only works in theory. In addition to having nearly a decade of time, they’ve raised $640M in capital, a number that when compared to the $9B market figure makes you wonder what Tekion’s investors are thinking. I suspect in another 5-10 years, Tekion will have sanded off the rough edges and it will become a stable, mature product. Time will tell. However, in the meantime, incumbent DMS companies will have had more than a decade to alter their offerings to compete with Tekion.

(Happy Tekion Customer)

Making something people want is incredibly hard. I don’t think this is appreciated in the finance or investment community (outside venture capital). To most investment professionals, companies are primarily numbers in the form of ROIC, valuation, etc. However if you just open your eyes at the sheer amount of death in startups and new product initiatives, you’d realize there are scores of very smart people who have worked 80-100 hour weeks, reasonable business thesis’ and years of their lives only to fail to make something people want. In finance speak: organic growth is difficult, especially for new products.

(Amazon of old)

Great software products are built on hard fought secrets. Secrets are deep seated insights about customers that customers themselves may not know. They’re the user flow, recommendation algorithms, user interface, the frictionless payment options and so on. What Reynolds spent decades on is what Tekion is spending hundreds of millions on: a deep, thorough understanding of the customer distilled into the product. Some of these secrets can only be gained through experimentation. When I was at Amazon, I talked to a fairly prominent designer who talked about the transition from Amazon’s old website to today’s. Amazon took forever to make the switch. The reason why it took Amazon years to transition was that although the new website looked better, the old website had a higher conversion rate and therefore Amazon would take a hit to revenue when the transition would have been made. Everything about the UI is and can be tweaked. I was told the specific font size and colour of the letters “In Stock” on a product page resulted in a single digit percentage change to Amazon’s revenue. As Amazon’s revenue is in the hundreds of billions a year, this change has made Amazon, quite literally, billions of dollars a year. Font/color changes take about 5 minutes to change. Not all changes are so easy. In the case of Tekion, it seems most of their challenges are in fundamental parts of their workflow and the reliability of the product. This indicates to me that not only are there product design issues, but potentially in the fundamental engineering architecture of the product. They have not identified and addressed enough customer secrets.

If a former Tesla CIO with hundreds of millions of dollars in funding takes a decade to create a reasonable product, what hope is there going after a small TAM product of which there is already a successful incumbent? Low. And that’s what makes VMS businesses wonderful. The thing about the quality of these businesses is, those secrets embedded in a product give management teams ample room to do stupid things. In a commodity business, there must be excellence in cost structure, channel, etc. In VMS, management teams can spend an infinite amount of money on R&D with zero to show for it. Why? Secrets are hard to come by. If you are interested in reading more on the topic, I’d recommend looking into “Design Thinking” or “Lean Startup”. In my opinion, the wonderful thing VMS investing opportunity affords us is value hidden in plain sight. An established VMS is a corpus of potentially durable secrets that can be had at a discount because management teams haven’t right sized the company.

Will AI Replace SaaS/VMS

On the topic of AI. Products are implemented in code and that seems to be a yellow flag for investors since AI seems to make code easier to write. As an example of something that happens in software development, the process of finding and addressing customer secrets is very much a manual process and probably will be for some time. Whether a company subscribes to design thinking, lean startup, or a more traditional product lifecycle process, someone has to go talk to customers. As evidenced by the graveyard of failed products, startups, and so on, this is very difficult. Product teams never have the complete psychology of the customer in mind. It is hard for a 20 year old Silicon Valley engineer to know the user journey of a 46 year old single mom in the midwest who works as a nurse, despite the fact that that mom might be the customer. Conversely, customers are not typically well versed in technology. It is unlikely for a 46 year old nurse to be able to craft user stories taking into account new Apple’s SDK features, projections on what percentage of customers have updated to a specific SDK version, scalability of eng architecture, how to comply with GDPR/CCPA laws, pricing models, user acquisition channels, etc. The product life cycle doesn’t just include customer interviews. It involves the rest of product design, product management, QA, data science, legal, finance, marketing, and so on. Each of those roles is a career with very distinct skill sets. At the time of writing, AI has been successful at automating none of the above. When I personally write code, as of early 2025, I still find it faster to write code unaided by an LLM. Perhaps in decades to come, some of these roles can be replaced by AI. From a finance standpoint, this probably wouldn’t be reflected in a DCF since DCFs that span generations basically make no sense to do. In the meantime, the more likely scenario is those jobs will be around for a long time to come, albeit aided by AI. Back when I started coding, there was no AWS, no Ruby on Rails, no Node.js, the web servers were not as sophisticated as today’s, and so on. Engineers back then needed to have some understanding of networking, memory management, etc. Today, less so. The tooling got better. As a result engineers can focus more on the business logic and less on the details. One way I think is worth thinking about AI is that it’s simply a tool akin to a web server, AWS, and so on. At best, all it does is allow easier expression of thought into code. At worst, it allows one bad engineer to produce the tech debt of ten good engineers.

Topicus

One business we have owned on and off since IPO is Topicus. Topicus is a serial acquirer of VMS businesses whose controlling shareholder is Constellation Software (CSU). It also trades at a very high LTM multiple so I'd argue it requires a fair amount of work to own. I’ve written about CSU in the past, but the general formula behind the business is to buy these durable VMS businesses at cheap prices, optimize the business for higher IRRs, then take those cash flows and buy more VMS businesses. CSU has returned ~25,000% since its IPO in 2006 so it’s a spectacular (potentially unmatched) business model. Topicus has been applying the CSU playbook, historically in Europe. Unfortunately, they’ve failed to ramp up acquisition spend in the past few years.

Every year, Topicus CEO, Robin van Poelje, seems to get picked on during CSU’s annual meeting. Some of it seems to be because Topicus has been lagging the parent company, CSU. In my imagination, for a high powered, ambitious CEO, this is fuel for the fire. It is likely humiliating to be called out by shareholders in front of your peers/board/etc. Last year, I started noticing Topicus M&A hires for Asia on LinkedIn. This is interesting because it seems likely that the competition from private equity et al. is likely lower. Furthermore, as Asia represents ~60% of the world’s population, the economies are less mature, and Asian economies are growing faster than in the west, the opportunity is vast. On January 16, 2025, Topicus announced the acquisition of PT Realta Chakradarma, an Indonesian (it’s first Indonesian acquisition) hospitality VMS. Furthermore, in 2025 Q1, Topicus has announced nearly more in acquisition spend than the past 5 years combined. As of the time of writing, Topicus has been trading at a constant price to sales multiple with a relatively constant look through margin. It stands to reason either Topicus has historically been expensive or it is cheap today. I think it is cheap today.

One large acquisition Topicus has announced is Asseco Poland at a total consideration of €417M. To contextualize the Asseco acquisition, it helps to look at another historical Topicus share acquisition, Sygnity. Sygnity has historically been a software consultancy with a large exposure to the Polish state. Topicus acquired its stake in Sygnity on March 22, 2022 for ~11.94 PLN/share via tender. At the time Sygnity had an EBITDA margin of 17.1%. Today, just 2 years later, it stands at 27.9%, a ~63% increase. Topicus installed its own management team and started acquiring companies, running the CSU playbook for inorganic growth. Today Sygnity trades at 76 PLN/share, representing a 637% gain in value for the shares and has a long runway for growth. It is worth mentioning we are shareholders in Topicus, Sygnity, Asseco Poland, and Asseco Poland’s subsidiary, Asseco South Eastern Europe representing ~39% of our portfolio.

Asseco Poland is a much more mature company than Sygnity and shares a fair amount with CSU/Topicus. It can claim to have more than 34,000 employees and has a long history of its own acquisitions. Historically, Asseco might grow cash flows HSD (~8-10%) where CSU might grow cash flows in the mid 20%s. Today, Asseco Poland has a ~15% EBITDA margin. It stands to reason that perhaps Topicus can replicate some of the operational magic it has applied at Sygnity to Asseco. However, as Asseco is a much larger organization that Sygnity is, I expect the transformation to take much longer. Furthermore, the quality of the company that Asseco has typically acquired is much higher than Topicus but that typically results in higher multiples being paid and therefore lower returns. As Topicus shares its operational and M&A playbook with Asseco, I expect both margin accretion and higher ROIC. As the share price of Asseco Poland indicates, the market is aware of how incredible the CSU playbook is. Topicus paid 85 PLN/share. As of the time of writing, Asseco Poland trades for 162.50 PLN/share indicating a tremendous amount of trust by the market. It’s also worth mentioning that CSU’s legendary founder, Mark Leonard, has had his eyes on Asseco for many years. Poland is a powerhouse of Europe being the fastest growing economy. However, the war in Ukraine broke out and investing in Poland was seen as a risk.

Since acquiring shares in Topicus (again), it seems American foreign policy has invigorated the animal spirits in Europe as Europe seeks to gain independence militarily and otherwise. In response to a stagnant economy and in the face of a more isolationist America, Germany has announced a €500M stimulus plan. This plan is equivalent to just under 1/10 of German GDP. I suspect going forward, we will see a more muscular Europe which likely sets Topicus and its subsidiaries up for a bright future.