Keeping Up With Data Overload: How Asset Managers Can Use AI to Optimize Efficiency and Performance
By Elizaveta Chistyakova
As the amount and complexity of data continue to evolve, an increasing number of asset managers are turning to artificial intelligence tools to augment and enhance their research capabilities.
If someone were asked to assign value to a Twitter post, many wouldn’t know where to begin. But we all remember a day in August 2018 when Elon Musk, Tesla’s founder, hit send on a harmless tweet, telling the world he intends to take his company private. Not only that, but he implied that the process was already in motion, by using what is now a famous phrase: “funding secured.” This single tweet from Elon, who has ~27 million Twitter followers, instantly sent the stock soaring 10% in record trading volume, briefly pushing Tesla above a $63 billion market cap. It didn’t take too long for the reality to settle in – the funding, in fact, was not secured, and the stock gave up the gain. One can only try to estimate the individual gains (and losses) of Tesla’s investors during this time.
The whipsaw in Tesla’s stock is an interesting case study – in a world of big data, stock prices can be moved by much more than just a company’s fundamentals, official filings, press releases or news reports. Investors and company operators now also have to track social media posts, chat boards and pretty much the entire internet. It’s a huge stream of data – Twitter alone generates 500 million posts a day. Musk in this instance is more of an exception than a rule – he has a massive follower base and his tweets about Tesla’s take-private were widely covered by the media; but how much information like this is flying under the radar, posted by those less known than Elon?
One way to manage this type of data at scale is sentiment analysis; a combination of natural language processing and text analytics applied to underlying data can provide investors with the ability to analyze publicly available information at mass and create a sentiment index, which then acts as a signal of potentially positive or negative impact. For example, a sentiment index can give an investor the ability to collect and analyze data from all customer review platforms, and conduct analysis on the overall perception of a product or service in the market, pick up on frequently mentioned words or phrases, and alert investors of a triggering event that can impact the overall value of the business in the future.
This is just one example of a broader theme – as the amount and complexity of data continue to evolve, an increasing number of asset managers are turning to artificial intelligence tools to augment and enhance their research capabilities.
AI adoption … The saga continues
Today, terms such as “big data” and “advanced analytics” are buzzwords we come across in all industries, including asset management and the world of finance more broadly; but while there is a lot of experimenting taking place, most firms (outside of early quant-driven adopters) are still in the early stages of wrapping their arms around what this new age of big data means for the investment world. TABB Group asked 160 buy-side firms about their use of AI and machine learning in 2018: two-thirds of them were at least exploring ways to incorporate AI and machine learning into their investment processes. Nearly all of them were planning to increase their spending in this area over the next 12 months. Hedge funds, especially those with quant-driven strategies, have been ahead of the curve. According to a survey from BarclayHedge, it turns out that more than half were already using AI or machine learning in their investment processes, a pretty notable 20% increase from the prior year.
While we are seemingly in the earlier stages of defining the AI phenomenon, the value and power of high-quality data are not to be argued. Within that, there is a category of “alternative data” – a term well-known to most in the investment world. This type of data is not necessarily produced directly by the underlying companies, but rather gathered through external sources, such as satellite data, internet searches, mobile data, credit card data or web traffic. The challenge is, however, that this type of data typically comes in an unstructured (or less structured) format. Data manipulation, to this degree, is outside of core skillsets available to most investment funds, and requires an unprecedented amount of computing power to collect, cleanse and store this data, as well as resources to figure out how to turn it into useful, digestible insights and incorporate these new datasets into existing investment processes.
This also requires a different type of talent, with skills different from those of a classic hedge fund analyst with a finance background, and pushes investment funds to expand recruiting efforts to former engineers and data scientists, who come with an understanding of data modeling and manipulation to a degree that most investors don’t. This leaves investment firms that don’t have deep pockets to hire a team of data scientists on the sidelines and not well-prepared for this new age of big data in investing, but not for lack of interest.
Adoption of unique alternative datasets in the asset management industry brings up the question of data value – in the rapidly moving investment world, alpha generation is key, and fund managers are willing to pay significant sums to chase it. However, as others pick up on the trends and purchase the same data, it naturally becomes less valuable.
The question then becomes: Why keep this data at all? The answer is simple – once it is widely adopted, it becomes table stakes. Imagine being at a poker table where all players but you know which cards will be on the river. Would you be able to place your bets to win the game? Another interesting side effect of wide adoption of specific datasets is that even if data is incorrect, but is available to enough fund managers with material positions in a given publicly traded company, they can still have the ability to move the market by reacting to the data they have, anticipating certain financial results or announcements and trading on this information–while neither may ever come, there is still an opportunity to gain.
The robots are coming … But you are safe (for now)
AI has created a lot of buzz, not just in financial services, but in every area and function, with the most extreme example of this being the emergence of virtual influencers. While not at all relevant to the world of finance, this certainly reminds us that no sector is safe. For now, however, the hysteria about AI eliminating jobs seems mostly overhyped, and the data quality that is out there to train the algorithms is simply not good enough for humans to take their hands off the wheel.
What AI is attempting to do, for now, is to enable us to be more efficient and make faster and better decisions with greater consistency by relying on technology. While this statement is not groundbreaking, it is worth reminding ourselves that unsupervised use of machines is still in the early innings, and if the fear of replacement can be set aside, together we can accomplish more and be more successful at our daily tasks.
A great example of this is Sentieo’s AI-powered platform, which enables investment professionals to more quickly and efficiently extract information from financial reports, earnings call transcripts, investor materials, press releases, and other publicly available documents through deep search and alerting functionality. Not only does this speed up the investment research and theses-generation process, it also allows teams to bring together data which is transmitted in all different formats and environments into a single workspace. It allows analysts to combine documents, add notes and save and share everything in one place to make the idea-generation process easier and more succinct.
There’s no question that the skill set of a portfolio manager is changing – simply finding information is not good enough. The barrier is going up for what it means to add value – analysis, and interpretation of the data remains on us, but it is also on us to learn and adjust types and formats in which this data is delivered. Big data will continue to shape the asset management space, integrating a broader spectrum of relevant information to empower better decision-making and save us time in the process. In a world with 500 million tweets being sent per day and the ability to instantly move markets, we simply can’t keep up with the information flow on our own – AI and machine learning tools can help us bridge the gap.
This article originally appeared on TabbFORUM.