Knowledge Graph Framework for Investment Decisions
Currently, research and trading decisions are mostly documented in linear formats: long, text-heavy documents that, due to their length and structure, can only present a limited slice of the market’s complexity and risk assessments. While knowledge graphs are already used to visualize entity relationships and ownership structures, they haven’t yet been applied to investment research (as of 2025), where their ability to reveal hidden connections could be transformative.
The mosaic theory is method financial analysts use to compile information from different sources to form a comprehensive picture of a given topic. The theory suggests that combining unobvious pieces of information can reveal a more complete understanding of a topics. Working in the financial industry I understand that time it takes to assess and aggregate countless data sources and formats - something AI is increasingly good at. I wish that the new batch of investment analysts entering the field can spend more time drawing connections and challenging base assumptions, instead. Thinking is the highest value task analysts ought to do. The knowledge graph framework fosters this lateral thinking and supercharges your ability to understand and dissect topics.
Application for Macro Research
To build out this system, the first step is to clean and organise your reports, data and notes into a centralised knowledge base, stored locally. Then you start categorising all your data, and finally you record connections between the topics.
The resulting output is a web of relationships that impact asset classes, surfacing hidden connections traditional methods miss. Graphing those nodes and connections creates a sort of 'virtual brain' of your team. This means every decision you make can be broken down into components, and tracked into its smallest 'particle'.
Key Components of a fully fledged Knowledge System:
- Entities: The “things” in markets (companies, policies, financial instruments, market players).
Entities are the building blocks of the graph.
- Processes: The actions or changes those entities undergo (a merger, a rate hike, or a credit downgrade)
Processes show how entities interact and evolve.
- Ontologies: The master blueprint that defines what entities exist, their properties, and how they relate (e.g., mapping how “liquidity” links to “systemic risk”)
Ontologies provide definitions and context.
- Semantic Layers: Thematic groupings that organize analysis and allow you to view the market through different lenses (Monetary layer, fiscal-real economy interface, investor type layer). Each layer hosts its own set of entities and relationships for focused analysis.
Host ontologies and provide structural framework for analysing interactions within and across domains.
- Triples: Connected/causal/correlative links like “earnings surprise → volatility spike.”
Triples are the fundamental way to express connections in the graph.
- Temporal Dimension: Tracks evolving patterns over time. (e.g., how the correlation between gold and equities shifts during different market regimes)
Transforms a knowledge graph from a static map into a living timeline
By organizing fragmented data into a network, this approach breaks barriers between teams and 'thought silos' – creating a collective reasoning tool that is fit for the age of AI.