The Filing Is Not the Intelligence
Every quarter, institutional investment managers with over $100 million in assets file Form 13F with the SEC. These filings list the equities they hold, the number of shares, and the fair market value of each position. The data is public, structured, and voluminous. In the most recent quarter, over 5,000 filers reported holdings totalling trillions of dollars across tens of thousands of securities.
Most analysts treat 13F filings as a scorecard: who bought what, who sold what, how did performance compare. This is a reading of the filing, not a use of it. The filing tells you what happened. The intelligence question is what it means—and what to do about it.
What 13F Data Actually Contains
A single 13F filing is a snapshot of a fund's long equity positions at a point in time. It does not include short positions, derivatives, cash, or non-US securities. It is filed 45 days after the quarter ends, which means the data is always historical. These are well-known limitations, and they are frequently cited as reasons to dismiss 13F analysis.
This dismissal is premature. The limitation of 13F data is not its lag or its incompleteness. The limitation is the analytical frame applied to it. When you read 13F filings as a list of holdings, the lag and incompleteness are fatal. When you read them as nodes in a network, they become something else entirely.
The Network Reading
Each 13F filing connects a fund to every company it holds. When you ingest thousands of filings into a graph structure, you build a bipartite network of funds and companies. From this network, you can derive relationships that the filings themselves never state:
Concentration analysis. Which companies are held by the most funds, and which are held by a small number? High concentration means broad consensus; low concentration means niche conviction. When the consensus shifts—when many funds exit the same company simultaneously—it signals a re-evaluation that often precedes price movement.
Cluster identification. Funds that hold similar portfolios are behaving as a cluster. Identifying these clusters reveals the implicit strategies driving the market: which funds are tracking the same benchmark, which are making concentrated bets, and which are diverging from their peer group.
Flow tracking. By comparing consecutive quarters, we can measure the direction and magnitude of capital flows between sectors, between companies, and between clusters. These flows are the connective tissue of market sentiment, and they reveal where institutional conviction is building and where it is fading.
Signal extraction. When a fund known for deep-value positions enters a growth stock, that is a signal. When a fund with a track record of early exits reduces a position by 40%, that is a signal. The filing does not flag these moments. The network does.
From Signal to Action
The distance between a signal and an action is where most intelligence products fail. They identify the pattern, describe it, and then leave the user to figure out what to do. This is the "interesting but not useful" problem, and it is endemic in financial intelligence.
Our approach closes that distance. When our pipeline processes a new batch of 13F filings, each signal is contextualised within the WorldGraph. A fund's new position is not just reported—it is connected to the company's supply chain, its competitive landscape, and its regulatory environment. A reduction in position is evaluated against the fund's historical behaviour, the behaviour of its peer cluster, and the company's network position. The output is not "Fund X reduced its position in Company Y by 30%." The output is "Company Y's supply chain shows two single-source dependencies that its peers have already begun diversifying, and three funds in the same cluster have reduced positions in the same quarter—this is a convergence signal, not an isolated trade."
That is actionable intelligence. It tells you what is happening, why it might be happening, and what the structural context is. It gives you something to evaluate and a basis for a decision.
The Pipeline
Building this pipeline required solving several engineering problems. The SEC provides filings as XML and HTML, not as structured data. Our ingestion system parses each filing, extracts the holdings, resolves each ticker to a company entity in the WorldGraph, and creates the corresponding edges. Entity resolution is the critical step: a filing may reference a company by ticker, by CUSIP, or by name, and each must be matched to the correct node in the graph.
Once the edges are created, the analysis layer runs automatically. Concentration metrics, cluster assignments, and flow calculations are updated. Signals are generated and prioritised based on the magnitude of the change, the significance of the fund, and the structural context of the company. The output is a prioritised feed of actionable items, ranked by relevance and supported by the underlying graph data.
The Standard for Filing Intelligence
SEC 13F filings are one of dozens of public data sources we process. They are valuable not because they are comprehensive, but because they are structured and reliable. When combined with the WorldGraph and the analytical layer that sits on top of it, they become something that a spreadsheet of holdings can never be: a window into the structural dynamics of institutional capital.
The filings have always been public. The intelligence has always been in them. What was missing was the infrastructure to extract it at scale and connect it to a model of the world that gives it meaning. That is what we built, and that is the difference between reading a filing and using one.