By Vinesh Jha, ExtractAlpha CEO and founder
Thank you to everyone who followed our research and alternative data journey at ExtractAlpha this past year! I thought I’d take this opportunity to review what we accomplished in 2023, and provide a sneak peek into what’s in store for the Year of the Dragon.
We launched two innovative new datasets in 2023. The first is a unique sentiment score based on the research reports of independent research providers (IRPs) in Asia, aggregated by our friends at SmartKarma. We’ve found this data to be highly predictive for developed APAC names, and it provides a differentiated and less biased view from what’s available from the sell side.
The second was the global expansion of our TrueBeats EPS and revenue surprise predictions, that we released for U.S. equities a couple of years ago. As many of you know, I’ve spent much of my career analyzing earnings surprises and analyst forecasts, going back to the founding team at StarMine in 1999, and TrueBeats was the culmination of a lot of that research. We’re pleased that some of our clients have enthusiastically started incorporating TrueBeats into their investment processes across global markets.
In terms of toolkits for investors, our Estimize platform now includes a widget that fintechs, media sites, and brokerage firms can embed into their websites, delivering differentiated, crowdsourced earnings estimates and enabling those platforms’ users to contribute their forecasts from outside of our own platform. Our widgets will be launching on several platforms soon.
Quant research is the lifeblood of what we do at ExtractAlpha, and we’ve continued to innovate by coming up with new use cases for interesting datasets. We released three major research papers in 2023:
- Using clickstream data on financial websites to identify high retail-attention names, and subsequently using retail attention to improve the performance of multiple signals, including classic anomalies like Value and Momentum. If you trade classic factors and would like to sharpen them, this research is for you
- Using TrueBeats to tilt portfolios towards industries and countries to improve asset allocation. This work was inspired by some of the models I worked on in the late ‘90s at Salomon Smith Barney. If you take views on industries or countries, this research is for you
- Using overlap in sell side analyst coverage to create new clusters of stocks, and then using these improved peer groupings to enhance classic technical signals such as momentum and reversal. If you run a stat arb book, this research is for you
As part of our research, we also track the performance of our datasets along multiple dimensions. We of course look at whether the signals we produce are effective in forecasting returns. By this metric, our Transcripts Model, which leverages our in-house NLP to score stocks based on the text of their earnings call transcripts, was massively successful for stock selection in 2023, generating a dollar-neutral Sharpe ratio over 3.2 before transaction costs (which are modest, given this is effectively a quarterly signal). All of our other U.S. equity signals produced incredibly positive returns as well.
We also track the accuracy of several of our datasets in predicting earnings and revenue surprise, namely Estimize, TrueBeats, and the Digital Revenue Signal, which uses consumer web activity to predict corporate revenues. We found that each of these produced accurate predictions 70 to 80% of the time, in line with historical averages. We track the hit rates live on the Estimize site. These measures give us confidence that our datasets continue to be fundamentally useful.
ExtractAlpha was also in attendance at multiple great events last year, including those hosted by Neudata, Eagle Alpha, and Battle of the Quants where I moderated data-focused panels in both London and Hong Kong. We also revamped our website. And finally, to toot our own horn, we won a couple of nice awards, including Best Alternative Data Provider to the Buy Side from WatersTechnology and Best Alternative Data Consolidator from A-Team’s TradingTech group.
Looking ahead to 2024, we are working on our product roadmap. We just announced a unique dataset offering with Japanese data. And, you’ll be seeing ExtractAlpha content in more places and consumable in more ways that will go beyond just quant trading, in particular leveraging new technologies to help discretionary managers ingest this data efficiently into their process; we believe strongly that investors of all stripes can benefit from our research. We’ll also be rolling out new signals based on some unique takes on news sentiment in multiple markets, as well as a big upcoming announcement in the ESG space.
Stay tuned and happy trading!