Using PAM to quickly form an opinion from large unstructured datasets
April 17, 2020 by Chandini
This case study shows how a user of the PAM platform is able to quickly investigate a new trade hypothesis by rapidly answering three main questions:
- What does the company ecosystem look like
- What are the macro-level takeaways for the company
- What are the specific areas I should focus my deeper analysis on
Whilst there have been many obvious consequences of the policy response to COVID-19 (increase telecommunication and suppression of the hospitality sector), these themes are accordingly well traded. For example shares in food delivery company, Ocado have over doubled in value this year, leading to it trading at a P/E ratio of over 17,000. Instead, more profitable trades tend to be made by identifying second or third-order consequences of an underlying theme - but this still requires you to act quickly.
“Failing to consider second- and third-order consequences is the cause of a lot of painfully bad decisions” -- Ray Dalio
There are lots of possible second-order interactions for any event and forming an opinion on a new name, or monitoring a new potential trade can be a considerable time investment. Just knowing what you need to know can be difficult in itself. This article will show how we can quickly form an opinion about a new name using Auquan’s PAM dashboard. We use a knowledge graph at the core of our products that provides a real-time understanding of companies and their ecosystems; employees, competitors, subsidiaries, industries, risks and more, allowing us to intelligently surface news that would otherwise take weeks of analyst research to identify.
Our example Idea: Duty-Free Retail (Dufry)
Duty-free retail is an industry that is one of these second-order effects that is worth investigating. Without air travel, all the shops that normally served these millions of passengers will likely stand empty. Within this theme, we’re going to look at Dufry which is a Swiss-based company operating shops in airports and cruise liners.
Understanding the company’s ecosystem
The first step is to bring up a quick summary of the company’s knowledge graph. Here we can see a summary of the company's ecosystem, including suppliers, competitors, personnel and risk factors etc. A snippet is seen below, with the numbers between entities indicating the relative importance of these connections (1 high, 4 low).
As we might have expected we can see lots of relationships to airports and similar companies. What might be more useful is the quick identification of subsidiaries that we may need to investigate further. The connection importance scores can change over time, in this case, air traffic has increased from being a low consideration to currently being of high importance. All these connections and strengths are mined by ingesting all public reports, filings and news about a company - allowing you to start seeing the big picture straight away.
Understanding the market consensus
Next, we can look at the news and other publicly available information about Dufry. The PAM dashboard summarises historic news by plotting the sentiment against price, with a risk indicator in the background. The sentiment was collated from 815,348 articles over the previous 3 months, it’s no substitute to reading article by article yourself, but it’s a hell of a lot quicker and can point an analyst to key dates or periods for further research. We can see here that sentiment started to fall, then after a while price followed this trend.
Taking a deeper look at news trends
Again, if we look one layer deeper, we can see that the volume of news was greater than in 2019, with several spikes that may be worth investigating further. The diffuse nature of the spikes (each spanning several months) indicates that this is unlikely to be causes by just a singular event, which would have been widely reported for a shorter period.
Taking the sentiment analysis another stage deeper we can see that not only was the overall sentiment negative but the distribution of articles was strongly positively skewed, further indicating a widely held negative public opinion.
We can dig further into this trend by investigating what themes are present in the news over this period. As we move through the first three months of the year we can see an increasing volume of news about coronavirus (purple). News about airports and transport that we would expect to be important for Dufry gets pushed aside and news reporting policy changes becomes more prevalent (Gold, 7.2% to 15.8%).
This can all be combined with other sources of data, such as the google trends for flights - as shown on the PAM dashboard below.
Results of our 10-minute overview
The final result is a quick 10-minute summary that has allowed yourself or one of your analysts to quickly understand:
- What does the company ecosystem look like: Including connections with other companies and sectors, risk factors and relevant themes, plus people and important locations
- What are the macro-level takeaways for the company: The falling market sentiment and share price, plus recent increases
- What are the specific areas I should focus my deeper analysis on: Spikes in sentiment and news volume coinciding with price moves could identify key drivers. Articles identified as mentioning corruption and crisis news articles should be explored further.
From here you are able to conduct deeper research to prove (or disprove) your hypotheses and stay on top of any new developments with minimal efforts. For example, here we might look into what news contributed to the leading drop in sentiment in Jan and how that compares to the increase in June. We might also look into the articles that were identified as talking about corruption in February or what caused the multiple news peaks in the last year. And so on.
Share this Article:
Auquan is a data science solutions provider for asset managers and hedge funds. Our state of the art technology empowers Portfolio Managers to stay ahead of the trend and achieve better returns.