Saga Partners Q4 2024 Commentary

Saga Partners commentary for the fourth quarter ended December 31, 2024.

During the second half of 2024, the Saga Portfolio (“the Portfolio”) increased 52.2% net of fees. This compares to the overall increase for the S&P 500 Index, including dividends, of 9.5%.

The cumulative return since inception on January 1, 2017, for the Saga Portfolio is 259.2% net of fees compared to the S&P 500 Index of 204.7%. The annualized return since inception for the Saga Portfolio is 17.3% net of fees compared to the S&P 500’s 14.9%. Please check your individual statement as specific account returns may vary depending on the timing of any contributions throughout the period.

Read more hedge fund letters here

(1) Saga Portfolio serves as a model for client accounts. Net returns assume 1.5% management fee.

(2) S&P 500 performance includes dividends. Please see disclaimer at the end of this letter regarding comparison to indices.

The Questionable Benefits of Frequent Updates and Interpretation of Results

This marks the 20th Saga Portfolio investor letter. When I started writing these letters in 2018, the goal was simple: to explain how I think about investing and to share what we own and why we own it. If someone else were managing my money, that is exactly what I would want to know.

That said, frequent updates can become distractions, shifting focus to short-term stock price movements rather than the fundamental drivers of long-term value. Most other investment strategies’ letters focus on what stock prices did in the past quarter. That is understandable—after all, the goal of investing is to grow wealth, and stock prices are the most immediate and visible way to track that progress. If the last few years have taught us anything, it’s that short-term stock price movements are unpredictable and rarely offer meaningful insight beyond the fact that more people either bought or sold shares during that period. A great business and a well-reasoned investment thesis typically do not change materially from year to year. As David Deutsch defines it, a good explanation is “hard to vary.”

The Saga Portfolio invests with an owner’s mindset because, in my view, that is the only rational way to invest. This naturally leads to low portfolio turnover. The best investments are typically left to compound over many years, provided they remain attractive, with changes only made when more compelling opportunities are discovered.

Given this approach and the fact that past investor letters have covered the Saga Portfolio’s philosophy in detail, I think an annual review of our core holdings is not only sufficient, but preferable. Going forward, I will write one annual investor letter after our core portfolio companies report their full-year fiscal results. However, I plan on sharing shorter, more frequent posts throughout the year on other topics that may be relevant or interesting.

Naturally, with less frequent portfolio updates and a consistent emphasis on focusing on the long-term, an important question is: Over what time frame should we judge investment performance?

In my view, one-year portfolio returns are largely random. Over a five-year period, assuming the starting and ending points aren’t unusual periods, we should generally expect to outperform the market. However, if the Portfolio hasn’t outperformed over a 10-year period and there isn’t a good explanation for the underperformance besides “the market is perpetually wrong,” then allocating long-term capital to a simple S&P 500 index fund is likely a better choice.

There’s an interesting paradox in long-term investing: while we ultimately expect our stocks to appreciate over time, true long-term owners and net savers should prefer lower share prices in the short term, all else being equal. It’s similar to homeownership: rising home prices may increase your net worth on paper, but if you plan to buy a second home, lower home prices are in your best interest. Likewise, as long-term investors, we should welcome lower stock prices as opportunities rather than fearing them as setbacks. How one reacts to price declines is a good indicator of whether they are thinking like a business owner or merely speculating—hoping to sell at a higher price to the next buyer.

There’s a saying: understanding the problem is half the solution. Beating the market over the long-term is challenging—especially when managing other people’s money. Few stocks outperform the market over the long-term and building a portfolio of numerous undervalued businesses is inherently difficult. Many of the best opportunities initially look controversial, some will be volatile, and a few will inevitably disappoint. This is why most mutual funds rely on bucketed investing strategies with wide diversification. While this approach makes them more marketable as they appear “less risky,” it has not been a recipe for long-term outperformance, with few exceptions.

In this letter, I explain the challenges of investing in publicly traded stocks through the lens of a concept discussed in the last investor letter—the growth of knowledge. This concept continues to shape how I invest and manage the Saga Portfolio. I’ve included my latest thoughts on our core holdings in the appendix. It has updates on Carvana, Roku, Trupanion, The Trade Desk, our latest investment in Wise PLC, and a postmortem on Redfin’s recently announced sale to Rocket Mortgage. I expect future annual letters will primarily focus on company updates and any portfolio changes, while general investing topics will be shared separately throughout the year.

Business Fragility and the Resilience of Knowledge

History suggests businesses are inherently fragile. Most startups fail, few companies last more than a decade, and even the strongest firms have eventually been replaced. Why is this the case, and what does it mean for investing?

Resilience is the ability of a system to maintain itself despite changes in its environment. It is hard to come by because the laws of physics do not inherently preserve structures beyond their basic elements. In The Science of Can and Can’t, Chiara Marletto describes how resilience emerges in certain systems through catalysts—systems that drive reliable transformations without themselves changing in the process. While planetary orbits and subatomic interactions persist due to the unchanging laws of physics, complex structures—like living organisms, machines, or businesses—require an explanation beyond physical laws to maintain their form. That explanation is knowledge.

Knowledge is a unique kind of catalyst that enables repeatable transformations. Unlike physical catalysts, knowledge is abstract—it can be stored, transferred, and instantiated across different mediums (e.g. DNA, brains, computers, books, etc.). Knowledge is information that is useful in solving a problem. Because it is useful, it tends to perpetuate itself once embedded in a physical system. In this way, knowledge is resilient and can make systems more resilient as well.

For example, the genetic instructions in a specie’s DNA can survive across generations, while an animal's ability to heal from an injury is a form of resilience that inanimate objects like rocks lack. Similarly, businesses embed knowledge into their products, processes, and strategies to enhance longevity—whether through rust-proof coatings on cars, quality control checks, or customer surveys for feedback. Without knowledge, these complex systems would not exist.

There are two known processes for creating knowledge:

1) Biological evolution, which occurs through genetic variation and natural selection

2) Human thought, which involves deliberate problem-solving through conjecture and criticism

Natural selection is slower and has limited reach because it is constrained by random mutations. Human-created knowledge, on the other hand, is explanatory, purposeful, and only limited by what is forbidden by the laws of physics. While no machine will ever travel faster than the speed of light, technological progress can continue to redefine what is possible, just as past societies would have viewed air travel or modern medicine as magic.

A corporation functions as an aggregator of resources, a more efficient way of coordinating people and capital to act as a catalyst for solving problems at scale. It facilitates repeatable transformations that create value. Take a car manufacturer: it starts with raw materials like steel, plastic, and rubber, applies a series of specialized processes, and produces a vehicle. What enables this transformation isn’t just the materials, but the embedded knowledge, the instructions, design, and know-how, that guide how those materials are shaped and assembled. It’s this combination of knowledge and resources that create wealth.

But if knowledge is resilient and businesses at their core are built on knowledge, why do most companies fail? The answer lies in two fundamental challenges:

1) Knowledge Can Be Copied

While difficult to create, knowledge is easy to replicate once created. Unlike biological evolution, where genetic adaptations spread only through reproduction, business knowledge can be copied instantly. In a free market, any profitable innovation attracts imitators, eroding excess returns. As a result, most businesses eventually earn just enough above subsistence levels to survive—until an economic shock or competitive shift wipes them out.

Companies can try to prevent their knowledge from being copied by:

• Legal barriers: Patents, copyrights, or regulatory protections (e.g. AT&T’s former government-sanctioned monopoly)
• Trade secrets: Proprietary knowledge kept confidential (e.g. Coca Cola’s secret formula)
• Continuous reinvestment: Using accumulated knowledge to maintain a lead (e.g. Intel’s semiconductor advancements)

Other competitive advantages—like economies of scale, network effects, and switching costs—can also create barriers to entry and make replication difficult.

Consider Microsoft Windows. It remains the dominant PC operating system due to a combination of these competitive advantages. However, it’s worth noting that despite Microsoft’s expertise and dominance in PCs, it failed to capture leadership in smartphone operating systems, where Apple’s iOS and Google’s Android now dominate. This highlights an important point: competitive advantages in one domain do not automatically transfer to another, even with significant resources and motivation.

2) The Creation of New Knowledge

In nature, if the environment changes too rapidly for a species to adapt, it becomes extinct. The same is true in business. Imagine a world where animals were able to create explanatory knowledge and purposefully modify their DNA to evolve. The pace of biological change would accelerate dramatically. Any species failing to adapt would become obsolete.

That hypothetical world is the free market economy, where companies must continuously keep up with competitors’ knowledge to survive. A stable environment favors incumbents, but rapid shifts—driven by new knowledge—can render once dominant firms obsolete. Just as evolutionary adaptations may become ineffective when ecosystems change, businesses can struggle when technological breakthroughs redefine their industries.

Consider newspapers. For decades, the leading paper in a city thrived due to high barriers to entry, including economies of scale and network effects. However, the rise of the Internet transformed the landscape, making their physical production (printing presses) and distribution (delivery trucks) assets obsolete. Newspapers were suddenly forced to compete for readers’ attention on a global scale against digital platforms like Google and Facebook, leaving many struggling to survive.

Businesses only persist by continuously solving problems better than alternative solutions. While competitive advantages may provide temporary protection, history is full of once-dominant companies that declined by clinging to old knowledge.

Take Henry Ford, who revolutionized auto manufacturing with the moving assembly line, making the Model T an enormous success. Ford’s efficiency, achieved through process refinements and vertical integration, gave it a massive cost advantage that other car manufacturers initially struggled to replicate. From the 1910s through the 1920s, Ford was one of the most profitable companies in America. However, Ford’s reluctance to innovate beyond the initial breakthrough allowed General Motors to catch up. GM adopted many of Ford’s techniques but also introduced flexible manufacturing, allowing for greater customization without disrupting production. As consumer preferences shifted toward variety and personalization, GM overtook Ford as the industry leader in the 1930s.

The fragility of businesses is not a flaw of the market, it is a natural consequence of how knowledge works. There is no law of nature that states a company must fail, yet many do because they fail to adapt to a changing environment. A company can have a long history of success and strong cash flows today, but it becomes fragile if it clings to outdated processes, chases short-term profits over long-term value creation, or fails to take promising new technologies seriously.

Competitive advantages provide temporary protection, but true resilience comes from having error correcting mechanisms—systems that allow a company to continuously refine its knowledge, improve its technology, and enhance its value to customers. Ultimately, resilience is not about avoiding problems, it’s about being best equipped to solve them when they inevitably arise.

The Challenge of Investing

While building a resilient business is hard, investing presents a different challenge. The average half-life of a publicly traded stock is ~10 years, meaning half of all companies listing in 2014 had disappeared from the market by 2024, often due to acquisitions or declining performance. Since 1980, over any given 10-year rolling period, around 85% of publicly traded stocks were either delisted or underperformed the S&P 500. Only about15% managed to outperform.

Note: The black dotted line represents the S&P 500’s performance. Orange and green columns represent relative under/outperformance relative to the S&P 500 over rolling 10-year periods. Data consists of stocks publicly traded stocks since 1980.

Source: Factset, Saga Partners

Why do most stocks underperform?

The primary reason many stocks underperform while only a few outperform lies in how the market tends to value businesses. In general, the market anchors a company’s valuation to its recent financial performance.

The job of the market is to evaluate opportunities and allocate resources efficiently. When it correctly assesses a business’s prospects, the stock price reflects its fair value, and is likely to generate average, market-like returns over its remaining life. This efficiency is beneficial for society because it directs capital to where it can be used most productively—to businesses solving problems people want most.

The challenge with assessing a business’s prospects is that the future depends on the creation of new knowledge, which is inherently unpredictable. If we knew the future knowledge that was to be created, then we would already have it available today. Therefore, it is impossible to know exactly what a company will earn in the future. Given this unpredictability, investors often rely on rules of thumb to value a company, the most common one being valuing companies based on recent financial results.

The problem with heuristics is that they are non-explanatory or are based on shallow explanations. Traditional valuation metrics extrapolate from recent results, assuming the past is a reliable predictor of the future. The philosopher Bertrand Russell illustrated the flaw in this kind of inductive reasoning with the story of a chicken: fed each morning, it comes to expect it will always be fed—until one day, it’s slaughtered. The chicken’s theory seemed to be supported right up until the day it didn’t. Similarly, while rules of thumb can be useful in certain contexts, they often rely on weak explanations rather than a deep understanding of a company’s long-term drivers of value.

Relying too heavily on past performance often overlooks the risk posed by future knowledge—new innovations or shifts that can render existing business models obsolete. This is one reason companies are frequently overvalued just before they are disrupted. If their initial valuations failed to anticipate the decline, they are likely to underperform the market. Even when investors recognize a company’s troubles—say, a struggling newspaper—and price the stock at a low multiple of recent earnings, it can still be overvalued if future prospects are worse than anticipated.

On the other hand, resilient companies that continue to innovate and solve increasingly important problems often defy traditional valuation metrics. A stock trading at 30 times earnings might seem expensive by conventional standards, but if the business has a long runway to scale a solution that is hard for others to replicate, it may in fact be undervalued.

Case Studies in Market Mispricing (Berkshire Hathaway, Walmart, Amazon)

The charts below show how each company’s stock performed relative to the S&P 500 over any 10-year forward period in its history, with the 10-year forward internal rate of return (IRR) shown on the left axis. This reflects whether the stock outperformed or underperformed the S&P 500 when held for the subsequent decade. On the right axis, a key financial metric is included to provide a comparison of the company’s fundamental performance over time.

What stands out is that many of the best-performing stocks historically outperformed during periods when traditional valuation metrics failed to capture their full long-term potential—and stopped outperforming once current results more accurately reflected future performance, often as reinvestment opportunities diminished and growth slowed.

Berkshire Hathaway’s historical performance illustrates this concept well. Book value per share, compounded at ~20% over six decades. However, most of its outperformance occurred from the 1960s to the 1980s, with returns aligning more closely to the S&P 500 after the year 2000, when its growth slowed.

Note: When the orange line is above the grey line, the company’s stock outperformed the market over the subsequent 10-year period. When the grey line is above the orange line, the company’s stock underperformed the market over the subsequent 10-year period.

Source: Factset, Saga Partners

Walmart became one of the most successful retailers in history by developing a superior discount retailing model built on operational efficiency and supply chain innovation. Yet in its earlier years, the market remained skeptical of its low-margin, high-volume approach. Throughout the 1970s and 1980s, Walmart’s stock was consistently undervalued as traditional metrics failed to capture the scale and durability of its advantage. It wasn’t until the late 1990s, when reinvestment opportunities waned and growth slowed—that valuation metrics began to align more closely with its intrinsic value. From then on, the stock largely tracked the broader market.

Note: When the orange line is above the grey line, the company’s stock outperformed the market over the subsequent 10-year period. When the grey line is above the orange line, the company’s stock underperformed the market over the subsequent 10-year period.

Source: Factset, Saga Partners

Amazon, the more recent retail (and now data center) giant, presents another compelling example. It discovered an improved retail concept with e-commerce and then scaled that service across the country. Traditional valuation metrics based on recent financial performance consistently underestimated its potential. As a result, Amazon had one of the highest returning stocks over the last 30 years.

Note: When the orange line is above the grey line, the company’s stock outperformed the market over the subsequent 10-year period. When the grey line is above the orange line, the company’s stock underperformed the market over the subsequent 10-year period.

Source: Factset, Saga Partners

It is also worth noting that despite their exceptional long-term returns, these stocks often endured significant volatility and steep drawdowns. Amazon, for instance, experienced regular declines of more than 50%, including a staggering 95% drop during its early years.

Source: Factset, Saga Partners

Exceptional long-term returns are rarely smooth; they are often littered with periods of intense market skepticism and steep declines. However, in many cases, these companies were still undervalued even before any major price drawdowns. Investors who were able to hold onto their shares throughout the volatility were rewarded with attractive long-term returns. Trying to trade in and out of a stock of a company that one thinks is significantly undervalued based on guessing short-term stock price movements is more likely to hurt rather than help results.

Sam Walton expressed a similar sentiment in his autobiography Made in America, “I believe the folks who have done the best with Wal-Mart stock are those who have studied the company, who have understood our strengths and our management approach, and who, like me, have just decided to invest with us for the long run.”

Conclusion

Many investors seek a universal magic formula or pattern to identify undervalued stocks. However, history shows that no single heuristic—whether based on valuation multiples, margins, capital intensity, or growth rates—reliably predicts long-term outperformance. The top-performing stocks of one decade had no greater probability of outperforming in the next.

That’s because rational investing isn’t about blindly applying heuristics or extrapolating past data. It’s not about deterministic cause-and-effect relationships. It’s about developing a good explanatory framework—one that begins with understanding the problems people are trying to solve, the range of solutions available, and how a company fits into that landscape. It involves studying how a business creates value, its resilience, competitive advantages, scalability, and why the market might be misjudging it. While predictions may be a byproduct of deep understanding, without a good explanatory framework, expectations about the future are meaningless.

More often than not, this process results in no actionable investment ideas because a compelling explanation fails to emerge. The investing process is about generating potential ideas and then trying to disprove them. The few ideas that can survive that scrutiny are the ones worth investing in. Even when a strong case for undervaluation exists, success is never guaranteed. Better explanations guide better decision-making, but certainty about the future is never attainable because it will always be inherently unpredictable.

This approach leads to a simple strategy: invest in the most compelling opportunities as if we had to own them for their remaining life. It shifts the focus from guessing short-term stock price movements to understanding the core drivers of long-term value creation. As a result, the Saga Portfolio often owns companies that might appear unconventional or even controversial at first glance, and may experience greater volatility (2022, anyone?). But over time, I expect these businesses, as a group, to compound value faster than the broader market.

As always, it is a privilege to manage your capital. The success of the Saga Portfolio depends on an investor base that is aligned and committed to the long-term—a mindset that is essential for navigating the market’s inevitable ups and downs.

Please don’t hesitate to reach out if you have any questions or comments.

Sincerely,

Joe Frankenfield

Appendix

Carvana

Carvana continues to demonstrate a rare combination of attractive qualities: a superior customer value proposition versus traditional dealerships, a complex and difficult to replicate infrastructure, industry-leading unit economics, and minimal risk of technological obsolescence. While the stock has recovered from its 2022 lows, the company still trades at a substantial discount to what I think is its long-term intrinsic value.

A key part of the original thesis was that Carvana has a structurally lower cost model—at scale. In the early years, the company was building out logistics and reconditioning infrastructure ahead of demand, which led to operating losses and skepticism about ultimate profitability. But detailed market-level cohort analysis offered visibility into its strong underlying margin potential.

By 2021, Carvana reached positive EBITDA, even as growth was prioritized over efficiency. Then 2022 provided a live stress-test of its model. A combination of logistical challenges from Omicron, affordability challenges driven by new vehicle shortages and rate hikes depressed demand, and the debt-funded acquisition of Adesa created a perfect storm raising serious concerns surrounding liquidity.

In response, management shifted focus from growth to efficiency, cutting $1 billion in annualized SG&A costs. Gross profit per unit rebounded in 2023 and reached new highs in 2024—all while maintaining high Net Promoter Scores—a testament to strong customer satisfaction.

In 2024, Carvana sold roughly the same number of retail units as in 2022, but with ~$900 million less in operating costs. Gross profit per unit reached $6,900, up from $3,000 in 2022 and $4,500 in 2021, resulting in $1.4 billion in EBITDA.

Source: Company filings, Saga Partners

EBITDA per unit reached an industry-leading $3,300. Retail unit volume grew by 33% year-over-year (50% in Q4 alone), all while utilizing just one-third of existing reconditioning capacity, suggesting even further fixed cost leverage.

Source: Company filings, Saga Partners

From 2012 to 2021, Carvana showed it could scale rapidly. In 2023, it showed it could be profitable without growth. Then in 2024, it demonstrated it can do both—achieving industry-leading margins while growing fast.

Still, some investors may question: is Carvana’s business inherently fragile? Could another macro shock send the company back into distress?

No business is immune to macro shocks, but 2022 was an unusually extreme environment. Historically, the used car market is relatively stable even during recessions. Today, Carvana is far better positioned to handle downturns. Prior to 2022, the company was essentially building the plane while flying it. Since then, its infrastructure and systems have matured, resulting in operational efficiencies that are now showing up in the numbers. Its margins per unit are double that of any other used car dealer, meaning Carvana can sell a similar car as CarMax and earn twice as much profit while providing a superior customer experience. This cost advantage not only drives superior economics in good times but also insulates the business relative to peers during more challenging environments.

The balance sheet has also strengthened meaningfully. At the end of 2024, Carvana held $3.6 billion in committed liquidity, which includes $1.7 billion in cash and an essentially untapped $1.9 billion floor plan facility. This compares to $5.5 billion in total debt, resulting in net debt of $3.8 billion—or 2.7x its 2024 EBITDA. Carvana’s two largest debt maturities are not until 2030 and 2031.

Crucially, management is committed to deleveraging. Although liquidity never fell below $1.4 billion, even at the depths of the 2022-2023 downturn, the challenging experience reinforced the importance of maintaining ample financial flexibility and never potentially needing to rely on external capital in times of stress.

Source: Company filings, Saga Partners

With shares recovering, is Carvana still undervalued?

In Q4’24, Carvana’s contribution margin per retail unit sold (gross profit excluding Adesa, minus variable operations and advertising expenses) was $4,765. This figure should remain relatively stable. Management intends to share most future gains with customers through lower prices, trade-in offers, lower shipping fees, etc. that will further strengthen its customer value proposition.

Carvana is currently only utilizing 35% of its 1.3 million unit annual reconditioning capacity. As volumes increase, fixed costs should remain relatively flat. For example, in Q4’24, retail units grew 50% year-over-year while SG&A remained flat (excluding one-time costs). At full utilization, a $4,765 contribution margin would provide over $6 billion in profit before fixed costs. Subtracting ~$1 billion in SG&A overhead, D&A, and stock-based compensation, and ~$600 million in interest expense, provides $4.6 billion in pre-tax income. Adding ~$200 million in Adesa profits provides ~$4.8 billion. Assuming a 25% tax rate would provide $3.6 billion in net profits, or a price-to-earnings of 10x its current market cap. Note with refinancing or debt reduction, interest expense could decline by another $200-300 million annually.

How fast can Carvana grow?

Growth is limited by inventory sourcing and reconditioning capacity—but unlike the past, the infrastructure is already in place. Inventory levels began ramping in Q2’24, with website units growing by 7,000-8,000 per quarter. If that pace continues and inventory turnover remains steady, Carvana would sell over 600,000 units in 2025 and 900,000 in 2026, and then fully utilize existing reconditioning capacity by 2027.

Expansion beyond that is also straightforward. Existing Adesa sites can be converted to support 3 million units annually with a relatively modest ~$1.5 billion investment. At current contribution margins, that scale could generate $14 billion in profit before fixed costs. The numbers become even more striking when considering if/when, and to what extent, Carvana scales past 3 million units.

Regardless of the exact timeline, the path is clear: Carvana has the business model and existing infrastructure to scale far beyond current volumes. As the company grows, both its economics and customer value proposition only get better. The “perfect storm” of 2022, didn’t break Carvana—it reshaped it into a more disciplined, more resilient organization, solidifying its position as the clear leader in online used car retail.

One can debate how much of the used car market will eventually shift online and Carvana share will be. But with a credible path to 1 million units—and 3 million not too far behind—even at recent high valuations, the opportunity looks highly attractive.

Roku

Roku continues to execute well in driving platform adoption, yet skepticism around its ability to monetize its platform and achieve sustained profitability remains. The investment thesis—that Roku will be one of the few end-state TV operating systems (OS), a position with substantial long-term economic value and profitability—seems simplistic but is worth unpacking given the market’s prevailing doubts about Roku’s long-term prospects.

The skepticism generally centers on two key questions: 1) Can Roku effectively compete with Big Tech, and 2) Even if Roku succeeds, is owning a TV OS ultimately a good business?

1) Can Roku Compete with Big Tech?

In many respects, Roku already has. Since spinning out of Netflix in 2007, Roku has remained singularly focused on building the best TV OS and doing so with a level of execution that Big Tech players have largely failed to match.

Historically, large and powerful incumbents often struggled in new markets outside their core business. This is not for lack of resources, but because they approached them from the perspective of benefiting their core business. While Google, Amazon, and Apple are formidable in their core markets, they built their TV OS platforms as extensions of mobile or computing ecosystems which require more powerful, and therefore more expensive, hardware. Their strategies approached the TV OS market as a way to serve broader strategic goals tied to their core business rather than best optimizing for the TV use case.

Roku, by contrast, was purpose-built for television. Its lightweight operating system requires less processing power, enabling OEM partners to manufacture lower-cost, more affordable smart TVs without compromising performance. Coupled with an intuitive and seamless user interface, has led to widespread adoption. In Q4’24, Roku-powered TVs outsold the next two leading TV OS platforms in the U.S combined, and Roku surpassed 90 million streaming households—now reaching more than half of all U.S. broadband homes.

Source: Company filings, Saga Partners

The question isn’t whether Roku can compete with Big Tech—it has for nearly two decades. I think the question more subject to debate is whether Roku’s large user base and market position will ultimately translate into a durable and profitable business.

2) Is Owning a TV Operating System a Good Business?

Building an operating system is inherently difficult, not necessarily due to writing the software, but because success requires broad ecosystem adoption by multiple different parties. In the case of TV, it requires integration with hardware makers, content providers, retailers, consumers, and advertisers—all of whom are reluctant to commit to a platform unless the others are already on board, creating a classic chicken-and-egg problem.

Despite these challenges, Roku has emerged as the clear leader in the U.S. TV OS market, with its platform continuing to expand internationally. Operating systems tend to consolidate around a few end-state winners due to network effects and scale advantages. Smaller, subscale players continue to exit or consolidate. For example, Vizio, which was the fifth largest TV OS in the U.S., was acquired by Walmart last year. Comcast’s X1 platform, despite having significant resources and built-in distribution, has struggled to gain meaningful traction. Meanwhile, TV OEMs like Sony, TCL, Philips, and Sharp, have increasingly moved toward licensing third-party OS platforms rather than developing and maintaining proprietary ones. Among OEMs, Samsung is perhaps the only player with the scale to viably maintain its own platform through its Tizen OS.

However, Roku’s dominance in the TV OS space has not yet translate into consistent, sustained profitability. That is to be expected. Building and scaling an OS is costly, with monetization generally following scale. Roku has deliberately prioritized expanding its footprint and developing its platform over near-term earnings.

Roku’s EBITDA fluctuations illustrate how the company can adjust its investment strategy based on market conditions. In 2021, revenue surged unexpectedly, prompting increased investment in the platform. However, as the advertising market weakened in 2022, Roku faced EBITDA losses. In response, management reduced costs in 2023, leading to a return to positive trailing-twelve month EBITDA. In 2024, Roku generated $260 million in EBITDA and is guiding toward generating operating profits in 2025.

Source: Company filings, Saga Partners

While profitability has fluctuated, Roku has made steady progress in monetizing its platform. In 2024, the Platform segment grew gross profit by 20%, reaching $1.9 billion.

Source: Company filings, Saga Partners

Connected TV (CTV) advertising is still in its early stages. While viewership has shifted from linear to streaming, ad budgets have been slower to follow, held back by legacy adverter buying practices. But as streaming continues to gain share and CTV advertising infrastructure improves around targeting and measurement, ad dollars will inevitably follow the eyeballs.

Roku represents a classic platform business—one that has secured dominant share in the world’s largest TV market while still expanding globally. As a result, it is still in the earlier, more capital-intensive phase of its lifecycle, which naturally weighs on current profitability. But as the industry consolidates around a few entrenched winners, those remaining platforms are well-positioned to generate meaningful, durable profits.

Trupanion

While many companies felt an immediate impact from COVID, Trupanion experienced a more delayed impact to its operations. From 2014 through 2021, the company consistently grew adjusted operating income at a high double-digit rate, reflecting strong execution as it scaled its product to new customers.

During the pandemic, pet owners delayed veterinary visits, lowering claims costs during that time. This led Trupanion to underestimate the rebound in vet visits coming out of COVID and therefore increased claims activity. Compounding the challenge, veterinary cost inflation accelerated faster than expected in the second half of 2022. The combination of rising claims and vet inflation compressed margins for the first time since Trupanion has been a public company. The trend continued into 2023, with higher-than-expected vet inflation leading to the first year-over-year decline in adjusted operating income in the Company’s history.

To address these challenges, Trupanion implemented price increases to restore margins. Since policy pricing for each pet is adjusted annually, the full impact of those adjustments takes 12-18 months to work through the customer base. As these changes took effect, margins began to recover in 2023, reached record levels by the end of 2024, with adjusted operating income increasing 40% year-over-year to $114 million.

Source: Company filings, Saga Partners

Note: Adjusted Operating income is a proxy for discretionary profit (pre-tax) Trupanion earns from existing pets before pet acquisition spend. It serves as a decent proxy for Trupanion’s value creation.

The auto insurance industry had a similar experience. During COVID, reduced driving led to fewer accidents and historically low claims ratios, prompting insurers to issue customer refunds. As driving habits normalized and inflation pressures drove up the cost of vehicles and repairs, claims ratios surged. In response, auto insurers implemented some of the largest premium increases in history to restore margins.

In insurance, inflation of the underlying risk often increases both the perceived value and cost of coverage—ultimately benefiting insurance providers. Reflecting this dynamic, Progressive and Allstate have seen their share prices reach all-time highs as rate increases fueled strong revenue growth and margin recovery.

Interestingly, Trupanion’s stock has not received the same credit. Despite a similar recovery in margins and revenue growth through price increases, its shares continue to trade at historically low valuation multiples. The market’s caution likely stems from concerns about future growth, particularly after new pet count growth slowed to 5% in 2024. Yet, even if Trupanion’s business did not grow at all, the current valuation still looks attractive.

At recent prices, shares are trading at 11x 2025 adjusted operating income guidance of $130 million. That means, if Trupanion chose not to reinvest capital in new customer acquisition and instead simply maintained its existing business, much of the adjusted operating income would drop to the bottom line. After paying interest expense and assuming a 25% tax rate, it would provide ~$90 million in net income, or 16x its market cap.

Of course, management doesn’t plan to stand still. Margin pressure in 2023 limited available funds for investing in new pet acquisition, which naturally weighed on growth in 2024. With margins and adjusted operating income now recovering, management plans to ramp up investment in pet acquisition. Given Trupanion’s strong value proposition (i.e. highest payout ratio in the industry), largely untapped market (with only 4% pet insurance penetration in the U.S. and Cananda), and plans to increase pet acquisition spend, I expect pet count growth to reaccelerate over the next year. That setup makes Trupanion’s current valuation even more compelling.

The Trade Desk

Since we first became owners of The Trade Desk in 2017, the company has delivered consistently strong execution. Despite the disruption caused by the COVID pandemic and broad macroeconomic volatility, The Trade Desk has scaled a global advertising platform while maintaining disciplined operational performance. Gross profit has compounded at a 35% annual rate over this period, reaching $1.9 billion in 2024.

Source: Company filings, Saga Partners

Note: Gross Profit is calculated as revenues less platform operations costs. This metric is used to show how the fundamental business has grown over time, not as a proxy for intrinsic value.

While the company’s long-term trajectory has been impressive, unexpectedly slower revenue growth of 22% in Q4’24, along with lower revenue guidance growth of 17% for Q1’25, led to questions of whether The Trade Desk’s opportunity and competitive position were weakening.

The core of the investment thesis is that programmatic advertising—leveraging data and software to target specific audiences more effectively—enables advertisers to reach their ideal customers with greater precision, measure performance more accurately, and optimize ad spending in real time. It brings data-driven decision-making and objectivity to an industry that historically relied on intuition and relationships.

The company’s long-term value creation hinges on its ability to capture a growing share of global ad dollars. Two trends support this:

1. Shift to programmatic advertising over the open internet as advertisers move away from walled gardens.
2. The Trade Desk’s platform will benefit as the leading independent demand-side platform (DSP) in an increasingly consolidated market.

A key distinction in the digital advertising landscape is between walled gardens and the open internet.

• Walled gardens (e.g., Google, Meta, and Amazon) own both the ad inventory and the platform required to purchase it. For example, buying ad inventory on YouTube requires advertisers to use Google’s DV360 demand side platform (DSP). Advertisers must operate within closed ecosystems, with limited transparency, no third-party verification, and restricted control over their data.
• The open internet, by contrast, provides access to a wide range of ad inventory across the web, mobile, connected TV, and audio media. This inventory is accessible through independent DSPs like The Trade Desk, which allow advertisers to buy media across platforms using their own data and measurement tools.

In the early days of digital advertising, walled gardens thrived due to their scale, superior identity graphs, and strong performance for specific use cases. However, walled gardens operate as closed, opaque systems that are not always aligned with the best interests of advertisers. The inherent limitations of walled gardens stem from their closed nature, restricting flexibility. Advertisers within walled gardens are forced to use the platform’s tools, reporting, and data policies. They platforms run closed auctions, set their own rules, value their own inventory, and measure their results, creating a conflict of interest. In effect, walled gardens, “grade their own homework.” This lack of transparency leaves advertisers unable to independently verify whether they are receiving the best value or if campaign results are accurate. As media consumption continues to fragments across devices and formats, the shortcomings of walled gardens have become increasingly evident.

In contrast, the open internet offers a wide array of publishers across the web, mobile apps, connected TV, and digital audio platforms. It provides advertisers with greater transparency, more objective measurement, clearer pricing, and reliable performance feedback. Unlike walled gardens, which are siloed, the open internet spans across multiple media channels and devices. With the advent of Unified ID 2.0, an open-source identity framework spearheaded by The Trade Desk, advertisers can increasingly target audiences across the open internet with greater precision, privacy, and interoperability. As a result, the open internet is well-positioned to capture a larger share of digital ad budgets.

In programmatic advertising, demand side platforms (DSPs) function as the control panel through which advertisers evaluate, manage, and purchase digital ad inventory across the internet. Trade Desk has emerged as the dominant independent DSP, acting as the backbone of this ecosystem. The platform benefits from network effects: advertisers want access to the widest range of inventory, while publishers want to be on the platform with the most demand. As more ad inventory flows through the platform, performance improves, drawing in more advertisers. This flywheel effect strengthens with scale, making it difficult for even a #2 DSP to achieve meaningful traction.

As an increasing share of ad dollars shifts to the open internet, The Trade Desk’s platform is positioned to benefit. Smaller walled gardens such as Netflix, Spotify, and Roku, have already opened their ad inventory to The Trade Desk recognizing that doing so aligns with their economic interests. By making their inventory accessible to a broader pool of demand, these platforms can generate more revenue. Over time, it will likely become in the best interest of the larger walled gardens to follow suit.

Yet, even if the largest walled gardens made their ad inventory available on the open internet, the opportunity remains huge. In 2024, global advertising surpassed $1 trillion, with ~$740 billion allocated to digital ads. Of that, ~70% went to walled gardens like Google, Meta Platforms, and Amazon, leaving ~$200 billion in digital ad spend across the open internet. In 2024, $12 billion in ad spend flowed through The Trade Desk’s platform, providing a long runway for growth within the open internet. While growth may experience fluctuations, such as slower growth in the fourth quarter that could extend into 2025, there is a strong case for ad dollars within the open internet to continue gravitating toward the Trade Desk’s platform.

Wise

In 2023, a friend and fellow investor introduced me to Wise PLC, a cross-border money transfer company he thought could be a great fit for the Saga Portfolio—and he was right. Wise’s mission to make international money transfers cheaper, faster, and more transparent immediately piqued my interest. The more I learned about the company, the more impressed I became. A deeper dive into the global remittance industry and the inner workings of the correspondent banking system only underscored the strength of Wise’s value proposition, the durability of its competitive advantage, and the scale of its long-term opportunity. It became less a question of if we would invest, and more a matter of when. Fortunately, the Portfolio had an opportunity to initiate a position last quarter.

Each year, consumers move ~£2 trillion across borders. About two-thirds of that volume flows through the traditional correspondent banking system—a complex web of bank relationships where large correspondent banks facilitate transfers between smaller local banks. These transactions are slow and expensive, often taking 2-5 days and incurring cumulative fees of 3-7% as each intermediary takes a cut.

Wise discovered a way around this inefficient system. Rather than relying on correspondent banks, it built its own global network of local bank accounts in each country, enabling it to receive and disburse funds domestically using local liquidity pools within each country. This eliminates the need for individual cross-border transactions. Money still crosses borders, but it does so on an aggregated, rather than a transaction-by-transaction, basis. Bank partners still earn fees for processing transactions when they are needed, but Wise is able to make transfers cheaper, faster, and more transparent.

However, even integrating with local banking partners comes with complications because Wise can’t control how quickly a partner bank responds to a transfer request and it can still be expensive. Therefore, Wise has taken further steps to bypass local bank partners and integrate directly with a country’s central payment system—the same “rails” used by banks themselves. By doing so, Wise gains complete control over the transfer process, delivering funds at the same speed and cost as a domestic bank.

In the UK, for example, Wise became the first non-bank to access the Faster Payments Service (FPS), eliminating ~90% of partner fees and reducing transfer times to under 20 seconds. In Hungry, connecting to the central bank cut costs by 14% and increased instant transfers from 17% to 82%. Since then, Wise has built similar integrations in Australia, Singapore, the Eurozone, and Brazil.

After 14 years of stitching together local banking relationships and direct connections to payment systems in over 160 countries, Wise processed £137 billion in calendar year 2024, making it the largest non-bank and the third-largest overall cross-border transfer company, trailing only JP Morgan and Citi, the two linchpins of the traditional correspondent banking network.

Note: Wise’s fiscal year ends 3/31, therefore 3Q25 is 12/31/24.

Source: Company filings, Saga Partners

These infrastructure investments are strategic. Establishing direct links to a country’s central rails can take years of navigating regulatory hurdles. Wise must prove it has the operational resilience, redundancy, and financial stability to meet the same standards as incumbent banks. It’s an upfront cost, but one that allows Wise to assume control of the full stack, replacing bank partners with its own infrastructure.

All of this supports Wise’s mission: to drive down international transfer fees as much as possible. It operates on a cost-plus model. Wise meticulously tracks its unit costs, adding a modest markup to fund operations and R&D, and targeting a 13-16% underlying income margin. As it scales payment volumes and achieves greater efficiency, Wise passes those savings back to customers in the form of lower prices. Its long-term edge stems from being the low-cost provider—driven by its relentless focus on infrastructure and operational excellence. This creates a moving target that’s increasingly difficult to compete against. Wise’s service becomes better and cheaper over time. The broader its network becomes, the more valuable it is to users—and the harder it is for competitors to replicate.

Wise’s main competition comes from traditional banks, but international money transfers are far from their core focus. Banks primarily earn fees by routing payments through the correspondent banking system, but their real business lies in deposit-taking and lending. Building a global money movement infrastructure from the ground up is well beyond their capabilities or strategic priorities. Legacy non-bank players like Western Union and MoneyGram remain reliant on costly, brick-and-mortar networks, making them less competitive on price and scalability. A new wave of digital competitors—such as Remitly, Revolut, and Currencycloud (now owned by Visa)—have entered the market, each with a different approach to cross-border transfers. However, a closer look at their unit economics and strategies highlights Wise’s operational discipline and relentless focus on cost efficiency as a meaningful competitive advantage.

By cutting out intermediaries and continuing to build its proprietary global infrastructure, Wise is able to consistently lower costs, improve service quality, and scale efficiently, enabling it to take share from the legacy correspondent banking system. In calendar year 2024, Wise processed $96 billion in consumer cross-border volume, roughly 5% of the $2 trillion total market. It’s $34 billion in small and mid-sized business volume accounted for just 0.5% of the $14 trillion market. These figures highlight the substantial runway Wise has to grow volumes and deepen its market penetration.

What makes Wise particularly compelling is that, unlike many scale-driven companies that often defer profitability in pursuit of growth, it is already meaningfully profitable. Wise generates strong free cash flow while growing at high double-digit rates—a rare and powerful combination that makes its current valuation especially attractive.

Redfin Postmortem

Following its Q4’24 results, Redfin announced it will be acquired by Rocket Mortgage for $1.75 billion ($12.50 per share) in an all-stock transaction. Following the announcement I sold our Redfin shares. This outcome was disappointing and ultimately resulted in an overall loss during our holding period.

What went wrong?

1) Entrenched Industry Dynamics and Relative Customer Value Proposition

Redfin’s mission has always been to make real estate transactions better and cheaper for customers. By leveraging its website traffic to generate leads and routing them to its agents, Redfin aimed to improve agent productivity and pass cost savings onto customers. This model stood apart from traditional brokerages and real estate portals. While Zillow and Realtor.com benefit from high commissions as lead generators, and traditional brokerages prioritize agents as their main customers, Redfin focused on reducing transaction costs for the benefit of homebuyers and sellers.

Despite offering a more cost-effective service, Redfin has struggled to achieve broader adoption. The underlying challenge stemmed from a deeply embedded system that insulated agent commissions from pricing pressure. Historically, multiple listing service (MLS) rules required sellers to pay the buyer’s agent commission, making the service appear “free” to buyer. Additionally, because this cost is typically wrapped into a 30-year mortgage, homebuyers remained less price-sensitive—even when Redfin could save them several thousand dollars.

This distorted incentive structure has made real estate notoriously difficult to disrupt. To use an analogy: it’s as if Uber tried to disrupt the taxi industry in a world where every driver is required to hire an agent to market their services—and, as part of that arrangement, must also pay the fees for the rider’s agent, even when the rider never requested one. Even though Redfin offered a similar (and in many ways better) service to traditional brokerages at a lower cost, many people simply didn’t place much value on the savings. The underlying thesis was that, as Redfin gained share and brand awareness, more homebuyers and sellers would recognize that they were receiving high-quality service without the premium price tag—rather than simply a “discount” or lower-tier experience.

Redfin’s value proposition, while real, was relatively incremental in terms of what customers truly valued. Compare this to a company like Carvana, where the difference in price, convenience, and selection compared to traditional used car dealers is meaningfully better. Redfin’s advantages were there, but there were subtler and less immediately apparent to the average customer, which made them less compelling and slower to resonate. Despite these structural headwinds, Redfin grew steadily throughout its history through 2021. It expanded into new markets, increased transactions and market share, and reached positive EBITDA in 2020—benefiting from strong housing demand and lower marketing spend following the pandemic. However, 2022 brought a shift…

Source: Company filings, Saga Partners

2) A Historic Housing Market Slump and High Fixed-Cost Business Model

Redfin’s salaried-agent model performed relatively well during periods home sales remained fairly stable. However, the sharp rise in mortgage rates in 2022 triggered one of the worst housing downturns in decades. Homeowners, no locked into low-rate mortgages, had little incentive to move, causing household turnover to plummet to multi-decade lows.

Source: Company filings, Saga Partners

When transactions tanked, Redfin’s more fixed-costed structure—paying agent with little to do—became a major liability. In response, the company cut costs and implemented layoffs to curb losses, but profitability continued to be out of reach.

Source: Company filings, Saga Partners

Why did I continue to hold shares despite weak operating results?

The most challenging question was determining whether Redfin’s struggles were a result of adverse industry conditions that would eventually improve or a broken business model. The housing market was enduring a historic downturn, but Redfin had executed well in the years leading up to 2022, and management had previously navigated tough period like the Global Financial Crisis.

The core model—leveraging online traffic to drive agent productivity—still made sense, assuming that, in time, people would return to buying and selling homes. Management appeared to grasp the challenges at hand and took steps to address them. The launch of Redfin Next, a new agent compensation plan that replaced salaries with traditional commission splits while retaining Redfin’s lead-generation support, seemed like a positive shift and improvement in their business model. It reduced fixed costs, making it easier and less risky to scale their agent base. Early indicators since the nationwide launch last fall have been promising, with stronger agent interest and hiring.

I thought the business would be able to navigate this downturn, but even in the case it ultimately failed, I still thought Redfin’s underlying assets were undervalued. Its website attracts ~50 million monthly active visitors and that level of organic traffic alone—along with its additional business segments like Rent, Mortgage, and Title— seemed worth far more than the company’s market cap since the stock’s decline in 2022. Consider these comps:

• Zillow acquired Trulia (with similar traffic and no other business segments) for $2.7 billion in 2014.
• Costar was rumored to offer $3 billion for Realtor.com (74 million MAUs) in 2023.
• Costar is acquiring Australia’s second-largest real estate portal, with significantly less traffic than Redfin—for $1.9 billion, while also spending nearly $1 billion in advertising annually to grow Homes.com traffic.

Compared to these valuations, Redfin’s $1.75 billion sale price seems low. I imagine Compass, with its emphasis on private listings would be very interested in owning the third largest real estate portal in the U.S. Alternatively, Redfin could have sold its rental business (which it acquired for $600 million in 2021), to unlock liquidity and buy more time to scale Redfin Next.

I continue to think that CEO Glenn Kelman remains earnestly committed to his lifelong mission to improving the real estate transaction. However, as Redfin’s liquidity options dwindled and the path forward became increasingly uncertain, he may have concluded that selling the company to Rocket Mortgage was the best way to keep the company intact. Unfortunately, this decision ultimately came at the expense of existing shareholders.

This experience will stick with me and is another reminder of how challenging it is to identify the true winners. I’ve closely followed Redfin and the broader real estate industry for years, and in hindsight, I overestimated the relative strength of Redfin’s customer value proposition while underestimating the difficulty of disrupting deeply entrenched industry structures. Perhaps Glenn Kelman’s compelling vision led me overlook the ongoing executional challenges the company has been experiencing.

While the outcome was disappointing, and the mistake entirely mine, there was a small silver lining: the acquisition announcement gave Redfin’s stock a lift just as the broader market was pulling back on tariff concerns. That created a timely opportunity to reallocate capital into portfolio holdings with stronger long-term potential. Of course, investing comes with its fair share of ups and downs, but it doesn’t make losses any easier to stomach when they do occur.