Matt Polyak had worked at Balyasny and Millennium before launching Hummingbird Capital. He knew what a large investment platform could provide. The advantage was partly human, but it was also mechanical: the ability to examine a portfolio from several angles without spending the morning assembling the evidence. In Arcana’s account of his experience, he put it simply: “It wasn’t just the people. It was the infrastructure.”
- Arcana connects portfolio construction, risk, performance, and research for institutional investors.
- Its central question: which returns came from your stock decisions, and which came from exposures shared with everyone else?
- Customers can test changes before trading and review the decisions afterward.
That distinction explains much of Arcana. A fund manager can have an intelligent opinion about every company in a portfolio and still misunderstand the portfolio itself. Different businesses can respond to the same market forces. A collection of carefully researched ideas can quietly become a wager on growth, interest rates, or a trade that too many other investors already own.
01 / The portfolio has its own opinion
Consider a hypothetical manager buying several companies for unrelated reasons. One has improving margins. Another has a promising product. A third looks cheap. Yet all three might carry similar exposure to a common factor. The company stories differ; part of the economic bet repeats. Counting names will not reveal the overlap.
Arcana makes software for examining that second story. Its performance attribution separates common factor contributions from idiosyncratic returns, the portion left after the model accounts for those factors. Its risk tools examine exposures, concentration, and stress scenarios. Crowding and ownership analysis add another question: who else may be leaning in the same direction?
return=Factor
contributions+Residual
return
The residual is what the model leaves unexplained. It does not, by itself, prove skill.
The useful word here is “model.” A residual is not a certificate of genius. It depends on what the model includes and how well it describes the portfolio. Arcana’s value is in making the questions inspectable: what drove the result, how much risk accompanied it, and whether the investment thesis explains what actually happened.
02 / From diagnosis to the next trade
Many systems can report a number. The harder task is carrying that number into the next decision. Arcana’s portfolio construction tools let managers compare a proposed portfolio with the existing one, examine a new position, change sizing, and screen possible hedges. An optimizer works against stated objectives and constraints.
That creates a practical sequence. Understand today’s exposures. Add the prospective trade. Inspect the change in risk. Compare alternatives. The attractive stock may still belong in the portfolio, but at a different size, or alongside a hedge. Research supplies a reason to own something; construction determines the shape of the ownership.
Arcana also supplies historical and live performance analysis, including earnings versus other trading days and the contribution of selection versus sizing. These distinctions make a postmortem more useful. Choosing the right company and allocating the right amount of capital are separate accomplishments. A profitable position can disguise a sizing mistake just as easily as a losing position can obscure a sound process.
03 / The spreadsheet assembly line
For Polyak, the appeal included the work that disappeared. Arcana’s published Hummingbird case study describes hours spent compiling investor information and manually entering data before adoption. The team connected portfolio data, reviewed exposures and attribution, and gained a place to investigate historical decisions. It is an account of reduced friction rather than proof of higher investment returns.
“It wasn’t just the people. It was the infrastructure.”Matt Polyak / Founder & CIO, Hummingbird Capital
A separate Arcana case study describes an unnamed $4 billion equity long/short fund consolidating analytics, risk, attribution, research management, and reporting. Arcana reports millions in savings, without a detailed public cost breakdown. The operational lesson is more portable than that number: count the human effort required to reconcile systems alongside the software invoices.
This is institutional software sold through a demo-led process. The buyer might be a portfolio manager seeking faster analysis, a risk team monitoring limits, or an operations executive simplifying reporting. Arcana’s economic argument combines analytical depth with fewer handoffs between people and applications.
04 / Keep the thesis attached
The research platform adds a different kind of memory. Notes, models, meetings, price targets, and conviction can sit beside positions and outcomes. Mock portfolios preserve recommendations without committing capital. Timestamping matters because an investment thesis can become suspiciously elegant after the result is known. A record gives the team something firmer to discuss.
The founders bring complementary experience. Rich Falk-Wallace, the CEO, previously worked at Citadel and Viking Global; CTO Siva Sooriyan held engineering leadership roles at LinkedIn. Arcana was founded in 2022. Their combination places investment practice beside the engineering problem of making a complex analytical system usable.
Its market position also resists a simple vendor-versus-vendor story. Arcana advertises more than 120 risk models, including models from MSCI Barra, Axioma, and Wolfe Research, alongside its own. Established model providers can supply part of the platform even as Arcana competes for the investor’s daily workflow. The differentiation rests on how the pieces connect.
05 / Take the context with you
The connections are extending outward. A February 2026 collaboration with CenterBook Partners brings analytics to its independent-manager network and priority evaluation for qualified Arcana users entering its alpha capture program. September brought an iOS app and an announced Hurricane Capital partnership. In October, Arcana announced its availability in Claude’s Connector Directory.
The Claude integration lets investors ask natural-language questions grounded in portfolio, risk, performance, and proprietary research data. It extends the same proposition into another interface: an answer becomes more useful when the system knows which portfolio, which thesis, and which history the question concerns.
The habit readers can copy is straightforward: record the thesis before the outcome, test a position in the context of the whole book, and review common exposures alongside individual names. Good data and appropriate models still matter; historical scenarios cannot anticipate every shock. Arcana gives institutional teams a way to practice those habits together. The next stock idea gets a tougher interview.