skip to Main Content
×Close search
Search
Quantitative Investing Blog - Lucena Research

Cracking the Code: Applying Alternative Data to Investment

On September 26th we were joined by a panel of industry experts for a discussion on applying alternative data to profitable investment.
Lucena Research

Erez Katz, Lucena Research CEO and Co-founder

On Thursday Sept. 26th, we hosted a panel discussion at the Grand Hyatt NY. The panel consisted of industry experts from both investment professionals and alternative data providers with the topic "Cracking the Code: Applying Alt Data and AI for Successful Investment". Moderated by Jonathan Neitzell, Founder, Anduril Partners LLC, our panelists were:

Brian Snyder, Sales Engineer, The Weather Company, an IBM Business

Peter McGinnis, Director, Equifax Capital Markets

Erez Katz, CEO and Co-Founder, Lucena Research

Elizabeth Pritchard, Founder, White Rock Data Solutions

Michael Beal, CEO, Data Capital Management

 

Content Highlights

  • Personal experiences and use cases for applying various types of alt data to investment.
  • A discussion on data pricing: what ranges are acceptable and how ready should the data be.
  • Is bias always a bad thing?
  • How the buy side can reduce friction in the buying process
  • Much more

Whether you are on the buy or sell side, the demands and interests in alt data are steadily increasing. The videos below will provide you with an overview of where the markets are heading and how to stay ahead of your competition. Enjoy!

 

Introduction and Keynote from Erez Katz 

 

Panel Content 

 

Panel Q&A

 

Additional Questions 

During the Q&A portion we received multiple questions that garnered in depth responses. The panel wanted to address each question fully, however due to time constraints we were unable to answer all the questions live. 

We were able to save the questions received but not addressed and would like to provide responses here. 

Q - Should ‘cost’ of alt data be surveyed in ‘bps’ of AUM? Assuming it has alpha, nominal dollar sets a barrier to entry for new vs. established.
A -
There are many innovative pricing themes in order to maximize revenue against over exposure. Pricing by Ticker, by AUM, by BPS are all used and some. The idea is to make the barrier to entry just high enough to extend some exclusivity to the high paying clients while maintaining fair market opportunity for all.

Q - How is global warming, a fact, affecting weather data based analysis?
A - Although we don't have a specific stance on Global Warming, we have over 30 years of climatology data and other additional historical data going back to 1979 to allow others to make their own decisions.

Q - As alt data becomes more widely recognized and adopted, will we see rationalization among the many current niche providers? Will there be enough pie for all?
A -
Our opinion is that there will be enough permutations of multiple orthogonal data sets for all. Price will determine supply / demand quotient to ensure the market dictates value / price relationships but there will be multiple levels of Alpha and Novelty driven offerings.

Q -
Suggestions for modeling time-series financial data?
A - This is a loaded question -- I recommend browsing our blog as we write extensively about alternative data validation and conditioning for deep learning research. There are many traditional and new technologies being used interchangeably where the winning approach has always been an Ensemble Expert voting. In Ensemble Expert voting we combine multiple uncorrelated models into a single decision engine. Also feel free to peruse our YouTube channel for some examples.

Q - Fundamental views on stocks tends from the last 3 months to several years, how long does most alt data live?
A -
Alternative data is not much different. Depending on the type of data and the research it can last anywhere from days to years. Lucena's focus has been mainly in the 1 day to 1 year range but that was mainly a choice we made. Others may have models with longer time horizon.

Q -
How do you deal with data sets that have different amounts of history? Can you use a 5 year data set with a 2 year data set?
A -
We normally look at the lowest common denominator for multiple factors with varied history. However we can always build new models and introduce them into the ensemble of models at the time their underlying data becomes more relevant

 

Additional Alt Data for Investment Resources:

Video: Finding Tradable Signals in Alternative Data

Video: Constructing Unique Data Feeds for Stock and KPI Forecasting

Blog Post: Live Portfolios from Alternative Data

Blog Post: Skepticism in the face of the Alternative Data revolution

 

Want to know more? Contact us or drop a comment below. 

More Posts

Never miss a new blog post, subscribe to our newsletter.