How Knowledge Graphs Can Help Asset Managers In 2020

There are many ways to define a knowledge graph. At its most basic, a knowledge graph is a large network that stores data on entities and on the relationships between these entities. These entities — which can be anything from concepts like “trade war” to real-world entities like an organization’s products, documents and data and their real-world relations — are stored in a graph.

Knowledge graphs can also be understood as a combination of a database (because structured queries can retrieve data), a graph (because the connections of the knowledge graph can be analyzed) and a knowledge base (because the interlinked data can be used to infer new facts). The ability of knowledge graphs to provide comprehensive views of vast sets of data and link these views to real-life connections will make the technology particularly useful for investment management in 2020 and the years to come.

2020 has been filled with extreme market events like the Covid-19 pandemic, which has caused countries around the world to shut down their economies for indefinite periods of time in order to limit the spread of the virus. Asset managers across the board have suffered historic losses, even while using market-neutral quant methodsthat worked during normal times, like taking equal long and short positions — many have questioned whether these are truly market-neutral. 2020 has been described as an “investing winter,” and fund managers have been urged to come up with new strategies that can weather the volatility. In these situations, a knowledge graph can aid in surviving market uncertainty.

Let’s pretend you’re an analyst researching a tire company. Without a knowledge graph, your research would require that you parse through any information you can find on a company, including its annual reports, investor presentations, earnings transcripts and bond prospectus. This could take hours, and you still might risk missing information.

With a knowledge graph, you could begin with a basic single view, which would display information on the tire company’s people, products, suppliers, competitors, patented technologies, trademarks and lawsuits and CSR activities. In that single view, the knowledge graph might point you to deep supply chain research, revealing that the tire company you’re researching uses a chemical supplier that’s dependent on a petroleum raw material exported by a company in Norway. Drawing from the news, the knowledge graph could show that in the next few months, there will be an increase in an export tax in Norway that would change pricing significantly. 

Using a knowledge graph, you could also quantify the risk that major new technologies pose to the tire company by looking at filings, NGO reports, government reports and patents to understand the strategies of key players. In 2020, an analyst might also use a knowledge graph to look at the risks the company might face from more complex news-related concepts, like the impact of the lockdown in Europe, the vaccine race, border restrictions in the U.S. and Amazon’s revised earnings forecast.

Graph representations also allow quants to look at companies by similarities, making it easier to cluster companies by their product, market or clients and unearth overlooked companies. For instance, at Auquan, we have created a comprehensive company knowledge graph with second- and third-degree relationships like suppliers of suppliers, people, entities, products, etc.

Knowledge graphs are not a new technology — they have been around for over 20 years. However, knowledge graphs have taken a backseat to other technologies because of their complexity. They’re difficult to implement and only make sense when one has vast amounts of interlinked data to make sense of.

However, recent advances in data extraction and natural language processing mean that a large volume of quantifiable and nonquantifiable information can be gathered from a variety of structured and unstructured datasets and texts. And compute power has increased a trillion-fold, which makes it easier to recreate these algorithms at scale.

Knowledge graphs and graph database technology are necessary to better understand data. For example, when Google started using knowledge graphs to enable semantic search in 2012, other technology companies like Facebook and Microsoft realized their power and started implementing them. Today, Facebook uses knowledge graphs to store all of our connections and make recommendations to users.

For investment management, knowledge graphs can act as a semantic search engine to find unexpected connections in investment research. They can extract millions of connections from anything an analyst would read, including company websites, filings, annual reports, news, sell-side research and databases. These connections can easily be stored in a knowledge graph and can produce new connections that wouldn’t be seen otherwise. Knowledge graphs can be the key to unlocking the complexities of 2020 and the years to come.

First published on

Similar Posts: