Sentieo’s AI-focused research platform supercharges equity analysts’ stock picking capabilities
In a world awash with data, hedge fund equity analysts looking for that one data point that could generate alpha face a significant challenge, akin to finding a needle in a haystack. But as data volumes have grown, so too has the pace of technology innovation.
There is now an impressive array of technology solutions to support the buy-side. One of these firms is Sentieo, a financial data platform utilising artificial intelligence and cutting edge search technology to essentially ‘Googlise” the equity research process; and in turn, help equity analysts find those metaphoric needles in haystacks by spending more time analysing, and less time searching.
Sentieo was set up by Alap Shah in 2016. Prior to this, Shah worked as a global equity analyst at Viking Global Investors and Citadel, two prominent US hedge funds. Over the course of his work at these institutions, Shah used Bloomberg and other market data tools to monitor the 150 or so global equities he covered on a daily basis.
“My job was to take the information I got, in whatever form, and try to turn it into a decision-making matrix, to build long and short positions in the portfolio,” says Shah (pictured).
“Most of the information coming to me was textural information or unstructured information. Some of it was financial data but the problem with that is that it is highly commoditised data, which reduces the ability to generate alpha.
“It is textural and unstructured data like filings, transcripts, broker research and company presentations that can give you an edge and allow you to see things others cannot see in the market. My job was to read that information and extract key data from it, and then try to structure it to use in portfolio models.”
During this period, 24/7 news coverage and the growth of social media platforms meant that these traditional tools used by Shah were suddenly becoming more limited. Trying to keep pace with the volume of data being produced was nothing short of attempting to swim upstream.
Cognisant of this pace of change, Shah decided to set up Sentieo, precisely to develop an AI-driven research platform that could industrialise the research process and equip equity analysts to work more efficiently in a data-rich world.
One of the foundational pieces of technology at Sentieo is the search function.
As Shah explains: “For two decades, people have used Google to get answers to their questions by putting in a few keystrokes. We are taking a page from that playbook and doing the same for financial research.”
To achieve this, Sentieo’s team started by creating a corpus of key documents that would matter for an equity analyst. Globally, there are approximately 50,000 global equities. Sentieo’s search engine runs by referencing some 15 million documents that correspond to those equities. Every SEC filing, every transcript from corporate conference calls, every piece of research from more than 1,000 brokers, every corporate presentation published, and every piece of relevant data from the news.
“All those documents live in the search engine and we are able to run not just a generic search but learn what the topics are that our users are researching to customise our services accordingly,” says Shah. “The system is able to learn based on user searches by recognising patterns using machine learning. If you search for the word ‘guidance’ – i.e. what is Tesla’s Model 3 guidance for Q3 – the system will also provide data on forecasts and expectations for Tesla.”
The premise can be summarised as follows: aggregate a comprehensive swathe of documents, put a search engine on top, and have that search engine understand context (relevant to the industry being researched) by utilising linguistic search algorithms. By using machine learning techniques, Sentieo’s search engine begins to work like Google to do things such as autocomplete (on keystrokes) and suggest search topics, based on what other users have found useful.
By doing this, it helps drive users to search for the things that are most important to them and, crucially, in a time-efficient manner. This is super-charged equity research when man and machine work to achieve an optimal outcome.
Shah explains that the way Sentieo is structured, there are three distinct pieces of the workflow.
The first element is the search workflow to look for and extract data. This then goes into the Sentieo Notebook, which is basically the nerve centre where the user aggregates all of their data (news items, CFO meeting notes, brokerage reports).
“If you’re the Tesla analyst in San Francisco and you have another analyst in Tokyo covering Toyota, and one in Germany covering Daimler – all of them will have their own notes, which you can see in the Sentieo Notebook,” says Shah, thereby facilitating collaboration across global equity teams.
“A user might start using Sentieo to search and consume documents. Then over time, as they get more comfortable with it, use it for data extraction, data cataloguing, and collaboration.
“The really important behaviour change, and something we focus on, is ‘Can I convince you to read most of the content that you care about, from a financial perspective, using Sentieo?’ Instead of cutting and pasting charts or research report extracts into Evernote, for example, you just click a button and everything is extracted into the Sentieo Notebook.”
This vastly reduces the time spent wading through the huge volumes of data to generate insights on specific stocks and gives analysts more time to construct a particular view on a company.
The third step of the workflow is a research management system. This is fed by all the information within the Sentieo Notebook.
If an analyst is recommending a BUY for Tesla, they can refer to all of the information they put together to support how they arrived at that decision: was it something the CFO told them last month? Was it something a Morgan Stanley report suggested?
“Everything that supports your investment thesis,” says Shah, “including your estimates on a company’s next quarter’s earnings or price target, for example, goes into the research management system.”
The CIO/CEO can then see everything documented in time, and what exactly drove the analyst’s’ decision-making process.
One might start by consuming raw data to build an investment thesis on a company. By extension, if everybody else in the team does the same for the other companies in the sector one is covering, one ends up with a database that has growth estimates and price targets for all the relevant stocks.
From there, the analyst can determine which stocks offer the biggest upside potential, which offer the biggest downside potential, in order to build the book.
The underlying AI technology of Sentieo’s platform means that it can funnel a vast amount of information, which it then distills into a concentrated mix, from which the analyst crystallises their BUY and SELL recommendations for the portfolio. It’s not quite alchemical but the aim, allegorically speaking, is to convert base lead (financial data) into gold (data points that can produce uncorrelated alpha).
The machine does the heavy lifting – it runs the searches, finds the most important topics, it does topic modelling to compare a stock’s KPIs over time, and builds all kinds of red flags. These might relate to legal changes that a company refers to in a transcript for the first time, or evidence that the CEO or CFO is skirting around the answer to important questions.
The more the system is used, the more the system learns what users do, allowing it to become more intelligent because of the network effect.
In respect to using alternative datasets, Shah observes that quant funds have their own black box models and will likely continue to follow this process, “but the space where we see the biggest innovation is in the fundamental equity research space”.
Fundamental analysts can’t be expected to reinvent the way they work overnight.
“Rather, you need to give them tools that cater to and understand their existing processes,” says Shah. “This particular alternative data set suggests this might be happening for this company – how do you figure out if this might be true and can be confirmed based on your existing fundamental research process?
“Sentieo helps fundamental investors contextualise the alternative data they are looking it, within the workflows they are used to. We are all former buy-side and sell-side analysts so we understand these workflow considerations. The system is built to take alternative data and push it through a client’s workflow to turn it into insights.”
Research platforms like Sentieo are finding ways to best utilise AI technology to augment human intelligence. The more Sentieo is used, the more it learns, helping analysts improve the performance of their stock picking to generate alpha in the portfolio.
“We think that the best humans plus the best machine can lead to 1 plus 1 equaling 3. And that’s what we are focused on scaling,” concludes Shah.
To listen to Alap Shah discuss the possibilities that Sentieo are exploring and what this means for hedge fund data analysts, register for the webinar: Using AI to supercharge equity research for alpha generation here