As thematic investing is gaining momentum, investors are confronted with the question of how to identify companies that are relevant to a specific theme. Data science company Entis is no stranger to answering these challenges. By bringing together their experience in artificial intelligence, structured and unstructured data, Entis helps to create actionable insights for investors.
When asset owners APG and PGGM were looking to translate their Sustainable Developments Investment (SDI) standard at scale, they turned to Entis. In July 2020, the SDI Asset Owner Platform was launched with asset owners APG, PGGM, BCI and AustralianSuper with Entis as and Qontigo as Global Distributor. The aim of the platform is to accelerate market adoption of SDIs, derived from the UN Sustainable Development Goals.
Wim Scheper, Ph.D., Managing Director at Entis explains what happens behind the scenes to produce the standardized SDI AOP data.
At the heart of the SDI analysis is our AI-powered data factory. By applying machine learning and natural language processing technologies, we are able to discover investable insights from large amounts of publicly available data. Important to note is that artificial intelligence is always combined with human intelligence for transparency, scale and continuous improvement of the SDI classification.
Our SDI operation consists of three main processes, in which the machine-human balance is carefully managed: Identify, Classify and Validate. These processes take us from the SDI taxonomy defined by the Design Authority (APG, PGGM, BCI and AustralianSuper) via a company longlist to a list of SDI qualified companies and ultimately, the SDI AOP Data.
High Level Process Flow of Our AI-Supported Methodology
Most of the more sophisticated technology is engaged in Step 2: Identify
As part of that process, large volumes of publicly available text are scraped from annual reports, financial reports, patents and company websites on thousands of companies across the globe. It is then mined, processed and stored in a database where natural language processing (NLP) is applied.
In a next step, this text is analyzed to determine which companies offer products or services that comply with the requirements for sustainable products derived from the SDI taxonomy owned by the Design Authority. This is done through semantic models that have the SDI taxonomy as their starting point. For example: ‘solar energy’ (SDG7) can be associated with: solar cells, photovoltaic, solar panels, PV-cells, etc. The whole of these terms is called a semantic model, which is used to determine what products a company offers. After having analyzed the texts of thousands of companies, the companies are ranked on the probability that they offer sustainable products.
Step 3 and 4: ‘The human touch’
Next are the Classify and Validate steps, which are where the ‘machine – human interaction’ becomes extra important. The SDI candidates (companies) resulting from the previous step, are presented to persons who classify the companies, i.e. map the company’s revenue to the SDGs. In the final step all data is checked through a series of automated checks followed by human validation. After validation the SDI data is ready for distribution.
Using the SDI AOP data in practice
Thanks in part to our technology, institutional investors can assess around 8,000 companies to find out whether they contribute (through their products and services) to the UN Sustainable Development Goals. There are many use cases for the data, for example, the data is used to report on the ‘SDI-ness’ of a given client portfolio or to build concentrated portfolios around one or more of the SDGs.
To learn how to access the data, contact email@example.com
To learn more about Entis, www.entis.ai