Alternative data will likely transform active investment management over the next five years, according to a white paper by Deloitte. Those firms that do not update their investment processes within that timeframe could, they argue, face strategic risks. Alternative data is a wide term that spans multiple categories. In brief, it refers to any non-traditional data (ie market price data, trade volume data) and includes online search data, trade data, satellite and weather data, consumer transaction data, geo-location data, etc.
“The amount of data is growing exponentially. IDC said that there were 16.3 zettabytes of information generated in 2017 alone (one zettabyte is 1 billion terrabytes). In order to process all that information and generate meaningful signals out of these vast pools of data, you need machine learning and cognitive computing solutions,” says Patrick Henry, Vice Chairman and US Investment Management Leader, Deloitte & Touche LLP.
In terms of alternative data adoption, the innovators and early adopters were mostly quantitative hedge funds seeking an information advantage. The MarketPsy Long-Short Fund, for example, first began feeding social media sentiment data into its investment models as far back as 2008.
According to Doug Dannemiller (pictured), Research Leader, Investment Management, Deloitte Center for Financial Services, the use of alternative data is becoming more mainstream and investment firms are increasing their technology and analytic capabilities: “Initially, alternative data and the analytics behind it were driven to tell you what was going to happen to a stock price at the next earnings call. Those picking up on the short-term bias of these signals were hedge funds, making short-term plays to try and gain an edge.
“Now, alternative data sets are capable of generating indicators of long-term business success and match with the investment processes of many more firms, including traditional long-only managers. They are able to support forecasts on a two-year or four-year time horizon, not just next quarter’s earnings call.”
For many years, company CEOs have hosted investment calls to discuss the latest earnings announcement. Sell-side analysts listened closely to look for additional data points to put into their models and determine what way the stock price would likely go. This was a traditional use of data to inform the investment decision.
But as Dannemiller explains: “If we think about alternative data, it’s the same investment call. However, now you have a machine listening in, using natural language processing, picking out key words to separate any false positive sentiment from the CEO from true conviction. The machine can correlate those key words across the industry, for multiple earnings calls, as well as against historical earnings calls from the same company. It puts things in a quantitative context rather than using the qualitative judgment of the equity analyst.”
One of the keys to success in this field is getting the right talent in place. It requires data scientists and engineers working alongside analysts and portfolio managers. To get the most value out of alternative data sets, Dannemiller says, “you need to have back and forth collaboration and iterative processes to know how to crunch the data to achieve a meaningful result”.
A portfolio manager might ask for the data to be twisted in such a way as to gain the necessary insight on a company. The data scientist then does back testing on the data set to see what it would have to done to the portfolio, historically, and forecast how it could impact the portfolio going forward.
“Then the investment firm can make changes in the investment decision model but it requires different skill-sets and people who work well together to achieve this,” concludes Dannemiller.
