StockSnips with Rebecca Wilde | E238

Using ML and NLP to make stock picks.

On today's Fintech Impact episode, Jason Pereira will talk to Rebecca Wilde, Managing Director of StockSnips. StockSnips is an AI and natural language processing platform that harvests data to provide stock purchase and sale indicators, bringing artificial intelligence to the world of stock selection. 

Episode Highlights

  • 0.38: At StockSnips we are providing investors with a low-cost high performing portfolio model that leverages our proprietary sentiment signals, allowing investment managers to go ahead and provide a competitive edge to their clients, says Rebecca.

  • 0.57: Rebecca has found a way through AI, natural language processing, and machine learning to derive a quantified signal that is a robust proxy for measuring investor sentiment.

  • 1.50: The goal was to figure out how can we apply AI to investing and what can be done to ultimately tackle the increasing volume of this unstructured data that's now available, says Rebecca.

  • 3.14: What Rebecca wanted to solve is how independent RAs can get back into the game of active management and not have to go down the rabbit hole of simply falling behind the passive indexes that have performed greatly.

  • 4.01: Rebecca talks about the type of unstructured data that they are harvesting in order to make the recommendations.

  • 6.20: Jason asks, assuming people start to figure out especially on their SEC filings and other things like that, what it is driving indication of sentiment and how do you adapt to that?

  • 6.54: We have message sentiment DK model that solves the problem of how you take raw sentiment data and create a signal from it, says Rebecca.

  • 8.03: Jason asks if someone hires or wants to work with you what do they get in return? Is it a simply a notification of what model portfolio or notifications of the buy and sell, or is there any kind of analytics provided in support of this? 

  • 11.08: Jason asks Rebecca, is there any kind of stories along the way as to something that you thought would be relevant, that turned out not to be relevant or something that really surprised you guys in terms of like we didn't think that would matter at all?

  • 12.13: It is a nice thing to see that all algorithms are picking up very accurately what is going on in the market and it's able to provide that up capture but also protect on the downside as well. 

  • 14.46: Rebecca talks about portfolio composition. How many positions do you typically hold for things and is there any allocation to cash?

  • 15.42: Right now, we just cover US equities, we cover all of them. However, in order to get strong enough signal, we typically have coverage for about 3000 smallest stocks that don't have enough media coverage, says Rebecca.

  • 19.32: If there was an easier way to reliably predict the style rotation, sector rotation, size rotation of markets we can be able to build much more robust models, says Rebecca.

  • 20.12: As per Rebecca the biggest challenge to taking company to till now has been creating a compelling value for customers and I think every firm faces this.

3 Key Points

  1. The cost of doing signal generation, getting the technology up to speed, and delivering it at scale is unfathomable for any individual investment manager to do on their own, says Rebecca.

  2. Rebecca has constructed a model with the wall street equity research firm where they have gone ahead and used growth value, quality, removed momentum and replaced it with a new sentiment signal and that has significantly outperformed many benchmarks.

  3. Rebecca is working not only on the education of artificial intelligence aspect but also how one can use this in its strategy or how can you use this in marketing to go ahead and gain new clients.

Tweetable Quotes

  • "It took several years for us to build sentiment signals by implementing the latest in natural language processing and training the machine structurally to be competent in financial literacy." - Rebecca

  • "We use machines to read over 50,000 news media articles a day and AI and machine learning algorithms pick up all of the relevant information and that allows us to construct a 360-degree view of a stock." – Rebecca

  • "We have three models right now that have been significantly back tested. We are touching about two years of live performance data and two of the models trade weekly and one of them trades monthly." – Rebecca

  • "You will find a lot of Robo advisors are taking very basic measures of sentiment and then making trades, which is moving markets and it is a big problem." – Rebecca

  • "We are kind of in our go to market approach. We have been targeting smaller independent RA's who have found great success." – Rebecca

  • "The fact that our models are entirely systematic, there is no human intervention, and it remains low cost. It leads to a very, very scalable model." – Rebecca

Resources Mentioned

Fintech ImpactJason PereiraFintech