The case for vertical software

Constellation Software, newspapers, and Japanese SaaS

The following post is not investment advice. Please do your own due diligence before making an investment.

If you enjoy this content, consider subscribing to be notified of any new posts. Thanks.

---

Returning home from the Berkshire meeting last Sunday, I had the opportunity to reflect about some themes I’ve been quite invested in. I’d like to share one of them: vertical software in Japan.

In 2021, I was a first year in our university's value investing club. Each week we would have classes on various topics ranging from moats to management to modelling. One day we did a case study on Constellation Software.

Constellation Software is a name that was not well known in Wall Street. Its founder, Mark Leonard, is just as elusive. A tall, bearded, Torontonian, you couldn't tell that this Gandalf looking man was running one of the most successful private equity companies in the world.

Constellation Software's strategy is simple: to acquire and become long-term owners of high-quality vertical software businesses.

Vertical software, or VMS, refers to software solutions tailored for a specific niche.

Take property management for example. In a horizontal software setup, property managers and owners have to log tenant data and track payment schedules on Excel, store lease documents and tax forms on Google Drive, and run their CRM on Salesforce. These products are general-purpose, and what happens is users in a specific field tend to bootstrap many of these to do day-to-day activities.

In a vertical software setup, property managers can do everything all in one place: Maintenance workflows can connect to vendor portals, recurring payments like rent can be collected and invoiced automatically, invoice and lease templates are set to abide by property-specific regulation.

--

There is plenty to admire about this type of business:

Competition is generally softer. Unlike extremely popular spaces like CRM or HR, a VMS in a specific field requires niche information. Typically, without personal experience or a strong connection to the space, this knowledge is hard to acquire. A property manager would know 10x more about property management than Bill Gates, and thus would have an edge in developing a property management solution. Ergo, it is quite difficult for large tech to enter a space with an established VMS. The smaller the field, the more prominent this effect becomes.

Secondly, and most importantly, is high switching costs, which gives the business pricing power. There are many takes on what creates high switching costs, but I think most of this can be distilled it into 3 factors, from most important: mission criticality, data embeddedness, and learning curve.

Mission criticality: To answer this question, one must simply ask, if the software were to shut down for a week, how much money or resources would the customer lose/forfeit? The Bloomberg terminal is by far the best example. In the high-pressure world of sales and trading, up-to-date data on securities and derivates is worth millions. One mistake, and the customer could potentially lose reputation and clients. Thus, if the Bloomberg terminal was shut down for even a day, their customers are in big trouble. If Bloomberg raised its prices 10% tomorrow, the trader would have no choice. This not only gives Bloomberg the ability to raise prices perpetually, but it also means its exceptionally recession-resistant.

Data embeddedness: There are many facets to this, but it simply relates to how long would it take to transfer all the necessary data from one software to another, and how stressful would that be. Quercus is a university-management software that, per customers, usually stores thousands upon thousands of student information, test documents, and course data. To transfer all that data to another software without disrupting school activities would be a gargantuan operation. You would have to get all students and teachers to transition. You would have to download and re-upload many terrabytes of data and files, which could take weeks if not months. And then after all that you have to hope and pray the new system handles this just as well as the old one. Yeah -- unless that product is defunct, that's not happening.

Learning curve: Not obvious on the surface but surprisingly important. Being used to how a system works creates trust. If you trust what you use, why even bother learning a new system. Adobe and Bloomberg are very common examples, since learning how to use them in itself could take years. But I would argue even simpler products like restaurant-management software have a deep learning curve. The sheer number of features necessary to operate a restaurant, Toast is an example, requires new users, who typically are not tech savvy, to have to learn how to use several products including order workflow, POS, inventory management, etc. The best vertical software strike a balance between usability and complexity.

I find the propensity of the 3 above-mentioned factors is most magnified when within the vertical the following are inherently present:

1. High regulatory standards - any field where government interaction is common often leads to high regulatory standards. Accounting is the most common: you do not want to make mistakes when presenting your financials to tax authorities. You also have certain fields which inherently have greater regulatory scrutiny, such as agriculture, healthcare, and energy. High regulatory standards raises mission-criticality and often data embeddedness when the regulation comes with high documentation requirements. This may, however, reduce learning curve as government filing and communications tend to be standardized, meaning different systems have more commonality.

2. Variety of business activities - more kinds of business activities means more complexity in software requirements. Hotels are a good example of this since it usually combines the activities of both a restaurant and a multi-unit property. Everyday, a hotel operators needs to accomplish a terrible variety of unique tasks ranging from logging bookings, check-ins, checkouts, calendaring the cleaning schedule, keeping records of maintenance issues, inspecting food deliveries, managing inventory, keeping track of orders, etc. These tasks all require separate products or portals, which increases learning curve and often data embeddedness.

3. High touchpoint with stakeholders - a business always has customers, suppliers, and employees, but what separates a high-touchpoint business from a low-touchpoint business is how often and how deeply they interact -- usually the case when a lot of coherent communication is required day-to-day. Universities and schools have extremely deep and recurring interactions with their customers and employees. This means both the school operator AND the students AND the teachers need to be on the software. Property managers have the same dynamic: both the landlord AND the tenant is on the platform to handle transactions and document transfer. While the non-core users: the students, teachers, or tenants, don't use the software as much as the core user, it does raise the data embeddedness as you simply have far more data and users on the system.

High-quality vertical software tends to have extremely low revenue attrition with high returns on capital and local monopolies or oligopolies. This allows them to continually raise ARPU through either 1) general price increases, or more effectively by 2) cross-selling new features. The latter has the benefit of additionally expanding product complexity and increasing switching cost, especially when these new features/modules are more niche or customized.

Another helpful note on examining a VMS is that, from what I’ve seen, successful vertical softwares are best a what’s called a “chokepoint.” This is a product or set of products within a VMS that provides a solution to the MOST mission critical aspect of running the business. For hotels, this is booking reservations. For property management, this is accounting. For hospitals, this is electronic medical record keeping. For manufacturing, this is inventory planning. Usually, finding the most common denominator across different competitors gives you more than enough evidence to determine this. But I mention this because it’s not always obvious which part of the product suite is the “chokepoint.”

What I’ve found is that it is far easier to a VMS to win by expanding to other products once they’ve established brand with their chokepoint product than for a VMS to expand into the chokepoint product once they’ve established themselves in other features.

—

What’s great about Constellation Software as a serial acquirer is that their M&A skills can scale. The playbook for a farm software will be relatively similar to an insurance software which is relatively similar to a healthcare software. As Constellation acquires more businesses, it can refine its models and KPIs, all while being able to connect knowledge and best practices among each vertical. Constellation Software's stock has seen excellent results, compounding close to 30% for the past 20 years!

In one of Leonard's letters, he notes the following:

"If Constellation had started in 1895 instead of 1995, we might have had the objective of being a great perpetual owner of daily newspapers. The newspaper industry underwent a long period of high growth which attracted many new entrants, followed by local consolidation, conglomeration, and eventual decline. I anticipate that the VMS industry will evolve similarly. "

The newspaper business and the vertical software business are not alike, but they both tend to achieve local monopolies if done right. While VMS exercised their local monopoly in terms of discipline, newspapers did so in terms of geography. Buffet was a huge fan of the newspaper business and invested in them heavily, including the Boston Globe, Buffalo Evening News, and most famously The Washington Post, which saw outsized returns compared to the S&P.

Customers wanted to read news that occurred locally and outside their city. This meant newspapers in a specific city had an inherent advantage on reporting stories within their geography, making it difficult for a "multi-city" newspaper even if they had greater capital. Furthermore, advertisers would prefer to work with the paper with the largest userbase. As a result, each city eventually had 1 or 2 dominant newspapers that only operated within the area and had significant pricing power. Buffet likened the newspaper business to the only bridge in a small town. No matter how much you hate it or like it, everybody pays their toll.

Buffet emphasized that the newspaper MUST be the dominant newspaper.

To see just how different a 2 vs 1 paper city is, we turn to the case of the Evening News. Buffet acquired Evening News for $32.5 million in 1976. Though the company was making under $2 million pre-tax annually, Buffet was expecting the growth and eventual monopolization of the Evening News to deliver far greater profits. The Evening News was the dominant paper in the sleepy steel town of Buffalo, having twice the circulation with its competitor, the Courier Express. There was, however, one weakness: they did not publish on Sundays whereas Courier Express did. Sundays were the becoming the prized day for advertising revenue as readers tended to stay on the pages longer. After Berkshire's acquisition, Buffet attempted to have Evening publish a Sunday paper -- Courier Express sued, accusing Evening of anti-trust. The bad press pulled away readers and advertisers, and Evening's own employees legally could not criticize Courier. By the end of 1977, the Evening News had just 1/4th of Courier Express's ad inches on the Sunday paper.

The papers were in all-out war. Courier Express had taken the opportunity to enhance their equipment, typesetting, and workforce. Evening News on the other hand focused on increasing news coverage. Eventually, Buffalo was struck with a recession, and both papers were losing money. By 1982, Evening News had accumulated a pre-tax loss of $12 million, despite somewhat closing readership gap on the Sunday paper.

Then came the light at the end of the tunnel. In September, Courier Express crumbled. While Berkshire, could sustain Evening with loads of cash to keep it going, Courier's parent company could not. Courier was in fact losing double the amount of Evening annually. It was a war of attrition, and Evening News had won. What was the result? Evening News had now become the paper in a 1 paper city. The difference? By the late 1980s, the Evening News would generate over $40 million in earnings annually.

—

Shedding light on VMS, I believe that switching costs alone is not enough. There also needs to be a future market dominance -- the paper in a 1 paper city. This is the risk of whitespace. Low penetration leads to a higher TAM, but that also means greater opportunity for competitors to erode dominance. Ergo, there has to be advantages to the current leading VMS that will allow it to continue leading as the market matures.

Bloomberg is by far the best example of this. When the Bloomberg terminal came on the market, they were up against Thomas Reuters, then the leader for financial data where you would think high switching costs would help maintain Reuters moat.

By building out thousands of features tailored to various needs, Bloomberg terminal customers were instantly hooked, and gave a large barrier of entry for new entrants. The final nail in the coffin was the chat features that provided a mission-critical network effect for traders to communicate with each other. Bloomberg’s switching cost and industry adoption became so widespread they gained near unlimited pricing power. They now charge north of $25,000 a head and continue to raise prices.

This brings me to my next topic, the case for vertical software in Japan:

As I've discussed in my older posts, Japan has been going through a cloud revolution. With a declining working population, incredibly low labor productivity, and prevalence of paper-based documentation, Japanese companies have been rapidly digitizing and bringing their data onto the cloud. I think this presents an incredible opportunity to find undervalued vertical software businesses.

Japanese companies have generally suffered from poor capital allocation. Many draw up excessive piles of cash that pulls down returns on capital. Coupled with a focus on profits over growth, and strict hierarchies, many Japanese stocks have failed to produce outsized returns for otherwise great businesses. This is changing as the government has been encouraging better corporate governance, such as encouraging stock buybacks and equity compensation. These factors, while not guaranteed to turn around the ship, remains a positive sign. Moreover, there has been more active M&A activity in Japan, such as the recent acquisition of Kaonavi by Carlyle or HRBrain by EQT, which I believe should return more value long-term.

There are several growing VMS businesses in Japan that trade below 10x EBITDA. Here are some that I own:

Property Databank is the leading property management software in Japan catered to large asset management firms, owners, and REITs. Trading at just 7x EBITDA, the company has been rapidly expanding its product capabilities outside of core rental property management and into building, retail, and construction management. Property Databank serves over half of all J-REITs. Churn is extremely low at <4% annually. Moreover, a significant chunk of revenue has come from its initial set-up and customization segment to support increasingly larger customer contracts. This has led to longer sales cycles and thus more volatility in operating profits. As the market has been overly concerned on near-term earnings, this has undoubtedly led to major sell-offs when project revenue recognition gets delayed, providing excellent entry opportunities for an otherwise excellent VMS business.

CYND is another interesting case. At 6x EBITDA, CYND has become the market leader in beauty parlor management software. With their key feature being a reservation booking system, CYND has expanded into website management, accounting, and e-commerce. The market punished CYND after they acquired their largest competitor, Kanzashi for a shit load of money. Kanzashi had a simpler product and a lower price-point, reducing ARPU. Furthermore, CYND depreciated the goodwill at a lifespan of 10 years, leading to excessive amortization expense which has killed bottom-line. I think however that CYND has never been at a better place. Echoing Evening News, CYND has taken over the number 2. Though they continue to face competition, they enjoy wide market leadership and extremely low churn of <4% annually. With the vast majority of beauty parlors unpenetrated, there is whitespace risk yet CYND has done an excellent job of ensuring their product remains the best in the market.

Both these businesses generate returns of capital north of 20%.

My main concerns are poorer capital allocation (most these businesses tend to have major cash piles and give out dividends when they should be reinvesting it back into the business) and the amount of time it may take for the market to recognize the quality of these businesses.

Despite this, this appears to be a unique situation with both an undervalued market and strong macro tailwinds. For the past several months, these are really the only businesses I’ve studied which I understand and can take conviction in. I see myself being long-term part owners of several excellent Japanese VMS businesses.

—