Repurposing Alternative Data for Various Investment Paradigms

  • By Paul Wilcox
  • July, 23
Blog Repurposing Alternative Data for Various Investment Paradigms

Erez Katz, CEO and Co-founder Lucena Research

What are some of the challenges data providers are currently facing?

Two main challenges are:

– Accessible and defensible evidence of their data’s value

  1. – Risk of overexposure

In the coming weeks our team will unveil several new relationships we have formed with major data providers seeking new ways to validate and commercialize their offerings.

In each of these engagements, we were able to provide empirical evidence of the underlying data’s predictive value through out-of-sample backtests, perpetually traded Model Portfolios, KPI forecasting reports, and Smart Data Feeds.

Here, I’d like to briefly describe how these reports and live models can be useful for both data providers and data consumers looking to maximize the value of alternative data.

Below is an example full life cycle of alternative data utilized by machine learning models in order to deliver multiple derived offerings. These offerings help providers and consumers overcome many of the challenges they face when using alternative data. 


Full life cycle of alternative data utilized by machine learning models


Alternative Data Adoption Challenges: 

For Data Providers

– Looking to enhance marketability and sales efforts with empirical evidence of data’s predictive capabilities.

– Discover opportunities for wider distribution and new audiences without lengthy trials/sales cycles or over-exposing raw data.

For Investment Professionals/Buy-Side

– A surplus of data opportunities.

– Ability to consume predictive data efficiently and affordably.


How Lucena Helps Both

We do all the heavy lifting, from data validation to advanced data science combined with machine learning technology to build predictive output. Investment professionals can easily  verify the output from our research and deploy the signals into their internal investment decision process. Here is how our derived offerings provide validation and empirical evidence of the data’s value.


What is a Model Portfolio?

Lucena delivers model portfolios algorithmically powered by big data and AI. The concept is as follows:

– Carry forward the very same execution rules of a backtest into the future where algorithmic decisions are made and published before the market opens. No human discretion involved.

– Simulate as authentically and realistically as possible in order to provide unbiased assessment of alternative data in real life scenarios.

– a. Account for slippage and transaction costs.
– b. Apply short borrowing cost when applicable.
– c. Account for dividends, splits, and reverse splits.
– d. Rich execution guidelines such as  OCO (order cancel order), stop loss, and target gain
allocation guidelines.

– Help define and validate new investment ideas with empirical evidence.

– Provide a comprehensive performance attribution report on demand for real-time assessment.

See Live Portfolios Using Alt Data 

Analyst consensus live portfolio

An excerpt from Lucena’s model portfolio collection. You can track performance in real time on the Lucena platform by following this link. Past performance is not indicative of future returns.


What is a Smart Data Feed?

A simple and cohesive daily output file consisting of buy/sell signals, ranking, confidence scores, and signal strength values derived from one or more datasets.

Signals are predicated on machine learning multi-factor models customizable to a specific asset universe and investment objective. The Smart Data Feed is delivered daily in an accessible and actionable format, easily incorporated into an internal investment decision process. 

Below is a glimpse at the Smart Data Feed data dictionary. Regardless of how many datasets or how many models were used to construct the signals, the output is always the very same format. Example of a Smart Data Feed brochure.


Buy/sell signals, ranking, confidence scores


How is Alternative Data Used for KPI Forecasting?

Multi-factor models based on one or more data sources can be very effective for KPI forecasting. Machine learning models suited to forecast a company’s gross sales or earnings per share turns out to be much more reliable relative to asset prices, mainly because KPIs are less susceptible to the noisiness of the market.

Here is an excerpt from a KPI forecasting presentation based on one of our data partners (announcement coming soon). As can be seen, a big contributor to forecasting gross sales for a retailer is seasonality and trend analysis. However, the alt-data and machine learning provide the critically important 7.5% excess accuracy residual.  


How is Alternative Data Used for Stock Forecasting?


Over the past five years, Lucena has amassed advanced technology suited for both data providers and data consumers alike. We are excited about the future and the opportunities the alternative data revolution will bring to our business and our partners.


Questions? Drop them below or contact us

Have a media inquiry or a topic you’d like to contribute to our blog?