This episode explains how Ryan Caldbeck uses algorithms to invest in consumer brands. He says the traditional metric 'velocity' (how fast products sell) has no proven link to success. Instead, he focuses on three things: store distribution, brand buzz on social media, and unique packaging. Key holdings: RX Bar (simple packaging, big exit), Halo Top (low funding, high valuation), and Liquid IV (his investment, revenue grew 8x in a year).
Ryan Caldbeck, in the podcast Invest Like the Best, explored how to introduce quantitative analysis into the private markets. Through the Helio system at his firm Circle Up, he applies quantitative screening to early-stage consumer brands. The research reveals that online sales vs. offline channel d
Ryan Caldbeck is the CEO of CircleUp, a firm that uses its proprietary system Helio to introduce quantitative analysis into private investments in early-stage consumer brands. The core thesis of this episode is: In the consumer retail sector, where business models are highly standardized, quantitative methods can effectively identify high-growth targets. However, deal execution in private markets still relies on relationship building, and this approach is difficult to replicate directly in the tech sector, where business models differ significantly. The most impactful insight in the entire episode is: Ryan Caldbeck argues that the metric most valued by traditional consumer investors—shelf velocity (i.e., unit sales per store per unit of time)—shows no verifiable correlation with future success across the tens of thousands of companies in CircleUp's dataset.
Ryan Caldbeck believes the starting point for quantitative investing in private markets is the realization that "manual screening can be systematized."
During his time in business school, Caldbeck’s job was to manually screen 400–500 consumer companies each week, relying solely on Google searches to determine whether they were worth pursuing. He found that "30 seconds was enough to get it right 85% of the time," which led him to think, "a computer could do it too." The initial Classifier model built by CircleUp aimed only to replicate the team's own "yes/no" decisions—inputting company financial data and outputting whether to accept it onto the platform.
After nine months of operation, the team uncovered a key insight: of the 10 variables most predictive of "whether to accept," 8 were external data points that did not require asking the company. This prompted the team to pivot toward building Sorcerer—an algorithm that proactively searches for companies and scrapes external information. Subsequently, the team upgraded the target variable from "whether we accept" to "whether the company succeeds," but due to the scarcity of exit data in private markets, they ultimately used revenue growth as a proxy indicator.
> Original quote: "We took a long time to pull together a large enough data set that we felt comfortable with that as an objective measure." — Meaning: The team spent a long time assembling a sufficiently large dataset to feel confident in using "success" as an objective metric.
Inference: Caldbeck points out that the core challenge for quantitative investing in private markets is the "long feedback loop"—exit data is sparse and success is difficult to define, making it necessary to find reliable proxy variables (such as revenue growth). This constraint makes model iteration speeds far slower than in public markets.
Ryan Caldbeck argues that the consumer retail sector possesses two structural features that make quantitative methods viable: highly uniform business models + vast amounts of accessible data.
Uniformity of Business Models: Consumer companies are essentially "manufacturing products and selling them," without involving the diverse models common in tech, such as freemium, subscriptions, or platform commissions. Caldbeck likens it to "playing the same chess game over and over again." This allows models to be reused across categories.
Data Abundance: Even the smallest consumer brand can access the following data through public channels:
Caldbeck emphasizes that this data "is almost the entire revenue equation for the brand," but the collection process is "extremely ugly"—requiring extensive cleaning, standardization, and merging.
Comparative Data: Caldbeck contrasts consumer retail with the tech sector—tech companies have vastly different business models (gaming vs. securities trading vs. cryptocurrency) and lack historical success cases as training data (e.g., when Uber emerged, there were no 100 similar targets to learn from).
Inference: Caldbeck believes that the penetration of quantitative methods in private markets will first occur in areas with standardized business models (consumer, real estate, some media), while the tech sector may never be systematically replaced—because even if an algorithm identifies the next Uber, it still needs to compete for deals with top VCs like Sequoia and Benchmark, and "I don't think they will be beaten."
Ryan Caldbeck points out that among hundreds of tested variables, the factors truly driving predictive power are concentrated in three dimensions: breadth and quality of distribution, brand intensity, and product uniqueness.
Counterintuitive finding: Shelf velocity — weekly sales per store — shows no correlation with success. Caldbeck describes this as "shocking" and cannot provide an economic intuition for it. He speculates the possible reason is that retail buyers prioritize "uniqueness" and "attracting new customer segments" over mere high turnover rates.
Inference: Caldbeck acknowledges that the relative importance of these factors varies with company size and category. The team finds that absolute revenue growth has a stronger predictive correlation than revenue growth rate.
Ryan Caldbeck argues that pricing in the private market can be data-driven, but trade execution still relies on relationship building. Quantification can only enhance transparency, not eliminate interpersonal factors.
Pricing Methodology: CircleUp uses revenue multiples for valuation, based on historical category data (e.g., a certain category averages 4x revenue and 200% growth), then adjusts according to the target's actual growth rate. Caldbeck acknowledges that this "still contains heuristic elements," and the team is building a rules-based system.
Case Study: Liquid IV (hydration drink mix brand)
Caldbeck compares private transactions to "buying a used car" — no matter the price, the buyer always feels "ripped off." CircleUp aims to make pricing feel "fair" through data transparency, much like Tesla or Apple: "This is the price range we are willing to pay, backed by this data."
Extrapolation: Caldbeck believes that quantification in the private market will spread faster than in the public market, because LPs and asset managers "crave scalable, repeatable investment strategies." However, the non-scalability of traditional top-tier VCs (such as Sequoia and Benchmark) is precisely their advantage — they cannot expand fund size indefinitely, but they can sustain high returns. The scalability of quantitative funds may attract large LPs, but they must first prove the repeatability of returns.
Ryan Caldbeck believes CircleUp faces two core risks: talent competition and building LP trust.
Talent Challenge: The team consists of "business professionals with financial backgrounds" and "engineers/data scientists." The former face high salary temptations from traditional finance, while the latter contend with an industry-wide talent shortage. Caldbeck states bluntly: "Nothing proves the need for reform in the U.S. education system more than trying to hire engineers and data scientists."
LP Trust: Unlike public market quant funds that can conduct backtesting, the private market has "longer feedback cycles and smaller backtesting data sets." Caldbeck’s selling point is that major consumer brands are being eroded by smaller brands (e.g., Halo Top’s sales surpassing Ben & Jerry’s), while traditional private equity (with $80 billion in assets) invests in "mid-tier brands" and fails to capture this trend. CircleUp is one of the few institutions capable of systematically investing in early-stage consumer brands.
Long-Term Vision: Caldbeck cites advice from investor Matt Christensen (son of Clayton Christensen), focusing on a 20-year vision. The company’s "constants" include:
Inference: Caldbeck believes the key barrier to quantizing the private market is not technology but "proving oneself." He references Bezos’ "constants" framework, arguing that CircleUp’s long-term advantage lies in its training dataset—the world’s largest financial dataset for consumer companies (tens of thousands of companies, accumulated over 6 years)—a moat that competitors will find difficult to replicate.
| Position | Guest Sentiment | Key Data |
|---|---|---|
| RX Bar | Positive case (product uniqueness) | Sold for $600M, raised only $10,000 |
| Halo Top | Positive case (brand uniqueness) | Valued at $1B, raised only $2M; sales surpassed Ben & Jerry's and Häagen-Dazs |
| Liquid IV | Positive case (investment target) | Revenue grew approximately 8x one year after investment; founder accepted a 50% lower valuation in exchange for data access |
| Beyond Meat | Positive mention (partner brand) | Announced IPO |
| Dollar Shave Club | Positive case (DTC exception) | Sold to Unilever for approximately $1B; $50,000 YouTube ad garnered 25M views |
| Vitamin Water | Background mention (brand strength change) | Consumer relationship weakened after acquisition by Coca-Cola |
1. "Shelf Velocity Has No Correlation with Success" (Ryan Caldbeck) — The metric most valued by traditional consumer investors shows no verifiable correlation across CircleUp's tens of thousands of company data points. The suspected reason: retail buyers prioritize "uniqueness" and "ability to attract new customer segments" over mere high turnover.
2. "The Fewer Words on the Front of the Package, the Faster the Company Grows" (Ryan Caldbeck) — Using computer vision to analyze approximately 2,000–3,000 snack bar brands, a negative correlation was found between word count and revenue growth. RX Bar is a classic example.
3. "Offline Channels Remain Far More Important Than Online" (Ryan Caldbeck) — DTC is suitable for product iteration testing but difficult to scale profitably. Evidence: RX Bar and Halo Top achieved massive success with near-zero funding, while many DTC companies raised $100 million+ and still remain unprofitable.
4. "Companies That Reply 'DM Me' on Twitter Are Three Times More Likely to Fail" (Ryan Caldbeck) — Because this typically indicates handling negative complaints and attempting to move conversations out of public view. It is a reverse signal of brand strength.
5. "Quantitative Methods Cannot Be Replicated in the Tech Sector" (Ryan Caldbeck) — For three reasons: inconsistent business models, lack of historical training data (e.g., when Uber emerged, N=1), and even if a good target is identified, one must compete with Sequoia/Benchmark for the deal — "I don't think they can be beaten."
6. "The Core Barrier to Quant in Private Markets Is Not Technology, but Proving Itself" (Ryan Caldbeck) — Public markets allow backtesting, while private markets have long feedback cycles and sparse data. CircleUp's selling point is capturing the structural trend of "large brands being eroded by small brands," which $80 billion in consumer private equity cannot achieve.
7. "The Most Valuable Asset Is Not a Specific Factor, but the Training Dataset" (Ryan Caldbeck) — CircleUp possesses the world's largest financial dataset for consumer companies (tens of thousands of companies, six years of accumulation), a moat that competitors find hard to replicate.
8. "Private Transactions Are Like Buying a Used Car — Quant Can Only Improve Transparency, Not Eliminate Human Factors" (Ryan Caldbeck) — CircleUp attempts to make pricing feel "fair" through data transparency, much like Tesla/Apple, but founders may still be persuaded by "tech circle friends" to accept absurd valuations of 20x revenue.