Sep 15, 2017
What happens when a buzzword like big data is uncovered as a half truth? While big data may be big in quantity, the data quality may be low grade and could even be hazardous to algorithms.
Factors that contribute to the declining value for much of big data include: fraud, inaccuracy, not representative of the market, incomplete, and one dimensional. It often requires unverifiable assumptions to be made for such issues as the meaning of a smiley emoji, or value of a “like.” Big data is also woefully void of the things that make a human “human,” such as their feelings toward the economy and retailers, and happiness in general. Big data falls short in the realm of future plans or purchase intentions and what motivates consumers. Even highly valued transaction data is still backward looking and could be holistically enhanced with real human data variables.
While the addition of the human element may seem as an apparent or requisite data input, the rush to employ technology to gather, scrape, and crunch anything other than real human responses has only served to create greater distance between much of big data’s consumer surrogates and real human beings.
However, two innovation firms, Consumer Edge Research and Prosper Insights & Analytics, intend to bridge the gap between big data short comings and real human insight to enable new predictive analytics for the financial services industry. Consumer Edge Research, a preeminent independent research boutique, will begin marketing Prosper’s unique, accurate, and predictive consumer datasets and analytics to the financial services industry. Prosper’s data was recently cited a top 25 source for big data and business intelligence.
An example of the types of new analytics/insights available is a higher order analytic that accurately forecasts the growth of GDP and also forecasted the 2008 economic collapse and recovery nine months in advance. The analytic was developed by Dr. Evangelos Simos, the co-founder of e-forecasting.com and professor of Economics at the University of New Hampshire.
This type of predictive analytic, and others such as a consumer sentiment forecast, would be valuable for those in the financial services industry and are developed from data “we don’t have” made by predictive modeling from “data we have.” It all begins with predictors built with factual data. An example of this is a macro input that identifies optimism or pessimism about future economic growth linked to the GDP (exhibit 7 below).
I recently posed several questions to Bill Pecoriello, CEO of Consumer Edge Research, about the impact that new, consumer-based data sources are having on the development of new analytics for the financial services industry.
Drenik: Bill, there is a great deal of discussion in the financial services industry about finding and utilizing non-traditional data to better understand things such as signal generation, portfolio optimization, risk management, etc. How do you see this alternative data sourcing movement playing out in the financial services industry?
Pecoriello: We are at the top of the first inning and asset managers are trying to figure out which data sets truly help them generate alpha versus adding to the noise. Also many are struggling with how to integrate the data into their investment process and what resources and organizational structures are required. I think we’re going to see a continued explosion of data sources in the near term and an eventual shakeout once investors are better able to qualify the data and separate truly high quality data from data that doesn’t improve portfolio performance. I also think the future will be about blending different data sources together to better predict the future versus just reporting on the past which will become commoditized.
Drenik: Recognizing that most data scientists realize that you need to start with good quality “alternative data” and let the models follow, what types of new predictive analytics can you envision for large quant shops in the financial services industry?
Pecoriello: I would break it into data that helps predict macro factors and data that is company specific. We are exciting to partner with Prosper as we see Prosper’s data helping investors make more accurate forecasts on both the macro economic as well as company specific fronts. On the macro front the Prosper data should help investors with improved GDP forecasts, consumer sentiment forecasts, and consumer spending forecasts. On the company specific front Prosper’s data can help investors improve sales forecasts for particular merchants and discretionary categories. Including linking the macro factors to the micro company specific factors driving sales growth. Alternative data needs to be more than alternative it needs to be high quality with a proven track record of generating increased alpha in asset managers portfolios.
Drenik: How might these alternative data sources be used or integrated with more traditional datasets that are currently available?
Pecoriello: At Consumer Edge Research we are providing investors with various transaction data sets such as credit cards and merchant receipts that do a great job of reporting what happened yesterday but the Prosper data will allow us to understand the “why” behind the transaction data as well as help predict future spending behavior. We see the blending of transaction data with the human element captured from Prosper’s data set resulting in better predictive tools for asset managers. That’s the future of alternative data for the financial services industry, combining several high quality data sets that result in better predictive tools.