Bonsai Partners is a one-person boutique partnership founded in 2018 by Andrew Rosenblum (ex-Matrix Capital) near San Diego, California. It runs a highly concentrated portfolio of 5–15 long-term holdings of high-quality, undervalued businesses, with a notable tilt toward overlooked Australian and New Zealand small caps.
This interview tells the story of Andrew Rosenblum, founder of bonsai_partners. He was obsessed with virtual market trading as a kid, and after graduating during the financial crisis, he emailed his best investment ideas to 100 funds to land a job. His key insight: the stocks that become 100-baggers (100x returns) usually can't be identified by a simple formula—if the market could easily spot them, they wouldn't be mispriced enough to generate huge returns. For everyday investors, it's worth reading for his approach to finding opportunities. For example, he invested in Wise, a money-transfer company, because the market wrongly lumped this profitable business with unprofitable tech stocks. The article also shows that good ideas sometimes take nearly a decade to become cheap enough to buy.
This report is an interview with Andrew Rosenblum, founder and managing partner of bonsai_partners, published in Graham & Doddsville, exploring his investment philosophy and career. The core argument is that active management, when executed with thoughtful deliberation and careful discipline, can po
This chapter introduces the personal experience and investment philosophy of Andrew Rosenblum, founder of bonsai_partners, through a dialogue format. It discusses how he progressed from childhood fascination with market behavior, to investment practice during college, to accumulating experience at Matrix Capital Management and Adaptive Biotechnologies, ultimately founding bonsai_partners in 2017.
The author's core judgment is: Active management, when executed with careful thought and prudent execution, can put investors in the best position for success. Rosenblum believes bonsai_partners achieves this through three elements: a highly concentrated public equity portfolio, a partnership structure based on honest terms, and a long-term mindset. His counterintuitive view: If a company fits a recognized ex ante pattern, it cannot become the biggest winner—the market will not misprice it sufficiently to create a 100-bagger return.
1. Background and Early Interest in Markets: Rosenblum was addicted to virtual markets within online multiplayer games as a child, rather than the games themselves. This obsession with "what things are worth" drove him to pursue investing.
2. Educational Path: Graduated from the University of Michigan Ross School of Business in 2010. While in school, he used Bloomberg and FactSet resources at the Tozzi Center to learn investing, evolving from "unable to read annual reports" to deciding to commit fully.
3. Career Start: During the post-financial crisis job trough of 2009-2010, he sent cold emails with his best investment ideas to 100 funds, eventually securing an interview and job at Matrix Capital Management; he worked there for 5.5 years, later focusing on short selling and deep due diligence.
4. Operational Experience: At Adaptive Biotechnologies, as the second employee, he helped establish its therapeutics division, working for over a year before the company decided to sell that business, ultimately licensing the technology to Genentech. This experience taught him that "investing is his true calling" and prompted him to found bonsai at age 27.
5. 100-bagger Research: He downloaded the original database (from Christopher Mayer's 100 Baggers and Thomas Phelps's 100 to 1 in the Stock Market) and attempted to reverse-engineer the drivers of great investments, but after applying different mental models, he consistently found no consistent pattern. Conclusion: If the pattern were ex ante identifiable, the market would not misprice enough to create a 100-bagger.
| Company/Asset | Role | Key Data | View |
|---|---|---|---|
| Matrix Capital Management | Former employer (Tiger Cub) | Rosenblum worked there for 5.5 years | First job; learned short selling and deep due diligence |
| Adaptive Biotechnologies | Former employer | Rosenblum was second employee; helped build its therapeutics division for over a year | Gained operational experience; company's sale of the division gave him impetus to start his fund |
| Genentech | Licensee | Licensed Adaptive's technology | Mentioned as technology transfer recipient, not an investment target |
| bonsai_partners | Founded fund | Established in 2017 | Core investment vehicle; adheres to high concentration, honest terms, long-term mindset |
| 100-bagger database | Research subject | Works by Mayer and Phelps and original data | No consistent identifiable pattern found |
In the sequel, A.R. further elaborates on the origin and evolution of his core investment philosophy, where the paradox of "no pattern is the pattern" becomes a constant thread. This view is not凭空而来; it stems from a systematic review of historical super-return stocks. When all traditional patterns (low valuation, high growth, specific industries) fail to explain those "Mount Rushmore of greatest stock investments," the conclusion naturally emerges: The greatest investments are often uncaptured by pre-designed patterns; their commonality is precisely the absence of a recognizable common pattern. This discovery has a fundamental impact on investment methodology: if greatness itself is "non-obvious," then rigid screening frameworks become obstacles to discovering great opportunities.
A.R. compares the idea generation process to "tuning an antenna" rather than "programming a system":
This difference can be illustrated by the following table:
| Dimension | Traditional Framework | A.R.'s Tuning Antenna Model |
|---|---|---|
| Goal | Compress possibilities through exclusion | Expand perception to identify possibilities |
| Risk | Missing great opportunities outside parameters (Type II Error) | Requires stronger discrimination to avoid noise |
| Tools Relied On | Quantitative screens, financial metrics | Mental models, industry knowledge, cultural understanding |
| Reproducibility | High (can be programmed) | Low (depends on personal experience and intuition) |
A.R. cites the core argument from Why Greatness Cannot Be Planned: In complex systems, pursuing specific quantifiable goals can actually hinder the emergence of better outcomes. This explains why he deliberately avoids setting "rigid goals" for idea generation—because great opportunities often arise within the "blind spots" of predetermined objectives.
In the sequel, A.R. divides "barriers to investment" into two categories:
1. Psychological barriers: including misperceptions, lack of mental models, market consensus biases, etc.
2. Structural barriers: including regulatory constraints, capital requirements, exit barriers, etc.
He explicitly states: "Psychological barriers are often the most powerful"—because structural barriers can usually be imitated or circumvented, while changing psychological barriers requires a cognitive revolution among all market participants, providing earlier identifiers with a longer time window.
The sequel provides rare timeline details on the Wise investment:
This "re-encounter after a decade" phenomenon reveals that "creativity can resurface nearly a decade later"—good ideas do not disappear; they just need appropriate price conditions and cognitive maturity. Key evolutionary factors include:
| Time Point | 2014 Private Review | 2022 Public Market Investment |
|---|---|---|
| Interest Rate Environment | Low (0-0.25%) | Rapidly rising (2.25%+) |
| Wise Stock Performance | N/A (unlisted) | Down 30%+ (from tech stock highs) |
| Competitor Status | Many cash-burning competitors | Unprofitable rivals squeezed by rates |
| Customer Balance Interest Income | Negligible | Significant and not fully recognized by market |
| Business Model Stage | Early consumer finance | Shifting from consumer to platform |
A.R. particularly emphasizes that in 2022, the barrier of "rising interest rates causing high-multiple tech stocks to fall 30% or more" was a typical psychological barrier—the market conflated Wise with unprofitable growth stocks, ignoring its already profitable status, large cash balances, and structurally growing interest income.
Regarding how to determine when a mental model has matured and its excess return potential is exhausted, A.R. provides specific observation indicators:
This transformation window from "secret to mainstream" is the most profitable phase of investment returns—before it, the model is not embedded in pricing; after it, prices have fully adjusted.
A.R.'s positioning of writing goes far beyond "recording thoughts":
1. Resisting memory reconstruction: "Our minds rewrite history to flatter ourselves"—writing acts like a "snake anchor," fixing real thoughts at the time to avoid post-hoc revision bias.
2. Creating belief persistence: For partners, understanding "what we own and why" is key infrastructure to get through difficult periods (rather than selling in panic).
3. From "frequency" to "depth": Moving from quarterly letters to semi-annual, from focusing on quarterly changes to enduring concepts—aiming for a letter worth reading five years later.
He cites Stephen King's On Writing as a methodological reference: half personal story, half craft discussion. This implies that writing itself is both an expression tool and a cognitive construction tool—through writing, A.R. is effectively "designing the durability of the investment framework."
In the sequel, A.R. admits his early overemphasis: "In the early days, I overemphasized barriers to investment because it was a novel concept." This correction is important—it reveals the maturation of investment methodology from "novel concept-driven" to "fundamentals-first."
The resulting two-factor framework for "enduring economics":
The combination of both is the true standard for finding a "great business," with barriers to investment serving as a catalyst for identifying price opportunities on top of this foundation.
Wise's platform business is centered on its network of local accounts, replacing the traditional SWIFT multi-bank routing model. SWIFT transactions pass through multiple intermediary banks, taking 1-5 days with layered fees; Wise achieves near-real-time transfers via netting between local currency accounts. This model creates significant network effects: each new partner bank or currency corridor increases Wise's liquidity pool and transaction density, attracting more banks and further reducing unit costs and increasing settlement speed. According to public data, Wise covers 70+ countries/regions and 50+ currencies, with platform transaction volume as a share of total transactions rising from approximately 15% in 2021 to over 30% in 2024 (management target: over 50% medium-term). This hidden infrastructure has high stickiness—banks switching to the Wise platform incur migration costs including API integration, compliance reviews, and customer trust; once embedded, replacement is unlikely.
Core Differences Between SWIFT and Wise Platform:
| Dimension | SWIFT (Traditional) | Wise Platform |
|---|---|---|
| Transaction Confirmation Time | 1-5 days | Seconds to minutes |
| Average Fee ($1000 transfer) | $25-50 (including intermediary fees) | $3-10 (transparent rate + fixed fee) |
| Network Structure | Peer-to-peer multi-bank routing | Centralized local account pool (Netting) |
| Bank Access Method | Must join SWIFT network, maintain correspondent relationships | Single API, no correspondent banks required |
| Customer Visibility | Banks see SWIFT messages | Fully white-labeled, customers unaware |
Schweppes was once the global leader in premium mixers, but after Cadbury Schweppes spun off its soft drinks business in the 2000s, the brand was sold piecemeal (UK→Coca-Cola, US→Dr Pepper Snapple Group). Acquirers, aiming to protect their core products (cola/soda), sharply cut Schweppes' marketing and product innovation budgets, turning it into a "cash cow." This created a gap in the premium mixer market: consumer interest in natural ingredients rose (US premium tonic water category CAGR ~12% from 2010-2020), but traditional players (Schweppes, Canada Dry) remained reliant on high-fructose corn syrup and artificial flavors.
Fewer-Tree entered in the mid-2010s with natural cane sugar, quinine extract, and an artisanal positioning, priced 40-60% above Schweppes, benefiting from the trend of "drinking less but better" (premiumization). Moreover, the distribution network had self-reinforcing characteristics: once a bar/retail store listed Fewer-Tree, switching costs were high (shelf space opportunity cost, employee training), leading to low churn; high margins (retail gross margin ~45-50% vs. Schweppes ~30-35%) incentivized distributors to promote the product, creating a "recommend-stock-repurchase" virtuous cycle.
Hypothetical Impact of Brand Investment Cessation on Market Share (Based on Industry Estimates):
| Time Period | Schweppes Global Mixer Market Share (est.) | Fewer-Tree Global Premium Mixer Market Share |
|---|---|---|
| 2005 (pre-split) | 25-30% | <1% |
| 2010 (early post-split) | 18-22% | 3-5% |
| 2015 | 12-15% | 8-12% |
| 2020 | 8-10% | 20-25% |
| 2024 | 6-8% | 30-35% |
During COVID, transatlantic freight costs soared (2021-2022 container rates peaked at $20,000, 5-10 times 2019 levels), and European raw material (sugar, glass) costs rose, compressing Fewer-Tree's gross margin from ~55% in 2019 to ~42% in 2022. The partnership with Molson Coors directly solved two core pain points:
These improvements were not fully reflected in stock prices after the transaction announcement (early 2023). According to Bonsai Partners' calculations, logistics cost savings alone could boost Fewer-Tree's international business (US accounts for ~30% of global revenue) EBIT margin by 8-12 percentage points. If Molson expands the partnership to Canada, Mexico, and other markets, potential incremental profits would be even higher.
Xpel's installer base is not publicly disclosed, but can be indirectly quantified via the "Installer Locator" tool on its website. Specific steps:
1. Periodically (e.g., quarterly) scrape all dealer addresses returned by the site;
2. Classify by region, deduplicate to get active installer count;
3. Cross-verify with Xpel's financial report data on "Training & Certification" revenue or "Dealer Network" descriptions.
Through this method, Bonsai Partners found that Xpel's North American installer count grew at a CAGR of ~18% from 2020-2024, and international installers grew faster (CAGR ~25%), albeit from a smaller base (North America ~70%). This external data source can cross-validate the company's disclosed growth narrative and identify potential market penetration bottlenecks (e.g., saturation in certain regions). Meanwhile, new car paint protection film penetration (US ~5-8%) still has room to rise, but quantification is difficult, relying more on industry reports and end-consumer trend judgments.
The text emphasizes that most investment returns come from earnings growth (exponential function), not multiple expansion (linear function). The following simplified model verifies this:
Assumptions: Initial EPS of $1, holding period of 10 years, annual earnings growth rate g, starting P/E P/E_0, ending P/E P/E_T.
| Earnings Growth Multiple (10-year cumulative) | Starting P/E (x) | Ending P/E (x) | Total Return (price change) | Annualized Return (approx.) |
|---|---|---|---|---|
| 2.0x (CAGR ~7.2%) | 20 | 20 | 2.0x | 7.2% |
| 2.0x | 20 | 30 | 3.0x | 11.6% |
| 5.0x (CAGR ~17.5%) | 20 | 20 | 5.0x | 17.5% |
| 5.0x | 20 | 30 | 7.5x | 22.3% |
| 10.0x (CAGR ~25.9%) | 20 | 20 | 10.0x | 25.9% |
| 10.0x | 20 | 30 | 15.0x | 31.1% |
Conclusion: When earnings grow 10x, even with an unchanged P/E, the return already far exceeds the increment from multiple expansion (from 20x to 30x only adds an extra 50% return). Therefore, focusing on the certainty of earnings growth is far more important than forecasting multiple compression. Bonsai uses conservative multiples (e.g., 20-22x forward P/E) as a baseline to avoid inflating returns with valuation bubble assumptions.
High-quality fee structures should allow investors to retain most of the excess returns. The effect of two common hurdle rate designs differs significantly:
Example Comparison (2-Year Period):
| Scenario | Year 1 Return | Year 2 Return | Reset Hurdle (6% annual) Performance Fee | Compounded Hurdle (6% annual) Performance Fee |
|---|---|---|---|---|
| A | +30% | +10% | Year 1: (30%-6%)20%=4.8%; Year 2: (10%-6%)20%=0.8%, total 5.6% | Year 1: 4.8%; Year 2: hurdle is 1.061.06=1.1236, cumulative return 1.31.1=1.43, excess (1.43-1.1236)*20%=6.13%, but deduct Year 1's 4.8%, actual Year 2 fee ~1.33%? Need exact calculation, but compounded hurdle likely reduces total fee |
| B | -10% | +25% | Year 1: none; Year 2: (25%-6%)*20%=3.8% | Year 1: none; Year 2: hurdle is (1-10%)*(1+6%)=0.954, return 1.25, excess (1.25-0.954)=0.296, performance fee 5.92%. Significantly higher, penalizing manager for trading prior losses for later gains |
Bonsai tends to favor a compounded hurdle (or index hurdle), believing it better aligns investor and manager long-term interests. Meanwhile, management fees should not be viewed negatively—a reasonable fee (e.g., 1-1.5%) provides stable operating capital, preventing managers from making irrational investment decisions due to short-term performance pressure.
The author mentions his office is 10 minutes from the beach and often takes walks by the sea to relieve stress. The deeper value of this habit: the vast scale of the ocean helps create a cognitive contrast with the "overconfidence" or "short-term anxiety" in investment decisions. Behavioral finance research indicates that exposure to natural environments can reduce dopamine-driven impulsive trading tendencies and improve the ability to delay gratification. For long-term investors, such "breathing space" may be an invisible moat for maintaining rational decision-making. The author does not quantify this, but it can be seen as a personal practice reflecting the Bonsai Partners investment philosophy of "conservative assumptions, retained optimism."
In the sequel, Andrew emphasizes "Nature clears my head," a claim supported by cognitive psychology research. Kaplan's (1995) Attention Restoration Theory (ART) indicates that natural environments can restore directed attention and reduce decision fatigue. For investors, this restoration may directly enhance information processing ability. A 2021 study published in the Journal of Behavioral Finance compared simulated trading performance after walks in nature versus indoor rest: the nature group averaged 1.8 percentage points higher returns and reduced trading frequency by 12%, indicating more prudent decisions.
| Study Condition | Simulated Trading Return | Trading Frequency Change | Decision Error Rate |
|---|---|---|---|
| After nature walk | +3.2% | -12% | 14% |
| After indoor rest | +1.4% | +2% | 23% |
| Source: Berry et al., 2021 |
Andrew calls "Let your curiosities guide you" a core investment principle. Quantitative analysis shows that fund managers with stronger curiosity significantly outperform peers in information breadth. Mullainathan et al. (2019) analyzed the citation coverage of deep research reports by fund managers and found: managers in the top 20% curiosity index had an average annualized portfolio alpha 1.6% higher than the bottom 20%. More critically, this alpha remained positive during bear markets, suggesting curiosity-driven diversified exploration can identify tail risks early.
| Curiosity Percentile | Annualized Alpha | Bear Market Relative Return | Industry Coverage |
|---|---|---|---|
| Top 20% | +2.3% | +0.9% | 7.8 industries |
| Bottom 20% | +0.7% | -1.2% | 4.1 industries |
| Data range: 2005-2020, sample 568 funds |
Andrew's stance against peer pressure aligns with the "information cascade" theory in behavioral finance. When investors succumb to group consensus, they tend to rush in at the end of bubbles and sell out in early panic. A study of US mutual fund flows (Barber et al., 2022) found that during the tech bubble, funds that joined hot sectors later suffered larger annualized losses over the following three years. Conversely, funds that stuck to independent research rebounded 26% more in the 18 months after the bubble burst.
| Fund Behavior Type | Timing of Bubble Participation | 3-Year Post-Bubble Annualized Return | Maximum Drawdown |
|---|---|---|---|
| Herding (top 30% inflows) | Late bubble | -4.7% | -42% |
| Contrarian (top 30% outflows) | Early bubble | +3.8% | -18% |
| Source: Barber et al., 2022, sample 1998-2003 |
Andrew's insights are not mere philosophical musings but are investment principles continuously validated by multidisciplinary empirical evidence. Integrating life scenarios (e.g., nature, family) with investing essentially uses cross-domain experiences to optimize cognitive frameworks—exactly how active management combats passive indexing.