Using Alt Data to Glimpse the Alt Future

Ted Cuzzillo, Principal Analyst – Data Analytics

Summary Bullets:

• Alternative data, or alt data, can indicate new trends much faster than conventional data

• Alt data is for domain experts only

The dawning of new trends might be described as being “like a string hanging down from the future.” That’s how the renowned forecaster Paul Saffo described it. The future first presents itself like the “odd event you can’t get out of your mind [that] could be a weak signal of a distant industry-disrupting S curve…”

Everyone’s looking for signs of the future. What’s going to happen to us? Who will suffer? When will the economy recover? Eyes are peeled for that weak but portentous S-curve.

The trouble, of course, is the abundance of those hanging strings. The great challenge is to distinguish strings that have real prescience. That’s true in ordinary times, but it’s far more important during upheavals like COVID and the shaken economy to follow. Alt-data customers say standard financial indicators and statistical releases are too slow.

Some hedge funds seem to be attempting such refinement. They are big clients for a relatively new industry that sells “alt data.” Among the nearly 1,500 alt-data vendors, business is booming. Hundreds of datasets offer data such as sentiment data on financial markets based on Twitter and StockTwits; web traffic data from company IP addresses to watch research topics by employees; and natural language processing (NLP) to analyze news, press releases, and financial social media.

Alt data, they say, detected the economy’s collapse much earlier than conventional data and, they hope, will be faster to detect any recovery. Indicators have included comparisons of social media posts in China with government statements to gauge the virus’s impact.

Doubters of “alt data’s” value complain that the signal to noise ratio is prohibitive. But to a savvy data analyst, noise only prompts another question, “What patterns in this noise can I find?” Doubters complain that the data’s more likely to be corrupt, but good data analysts always stay alert for that.

Even very noisy data can pose an odd event you can’t get out of your mind, as Saffo put it. You may recall, possibly far in your past, a visit to a storefront psychic. She gazed into her big glass orb, then up at you with a knowing gaze she’d perfected over decades to give you advice she gave to everyone. It was all quite a show. But even so, you still couldn’t shake one thing she said that day even though you forgot everything else she said.

The odd event persists whether it’s from a fortune teller, a movie, or an overheard remark. The odd outlier on a chart may also persist and beg to be explored for verification or rejection.

Whether the data is conventional or “alt,” it’s subject to human judgment. The choice is not either to abandon “alt data” in favor of strictly conventional data nor to embrace it with abandon. Alt data can be used, but only when backed by domain knowledge that helps to know when a ringing in the ear has real meaning.

Like so many other powerful tools, alt data and conventional data can be used to great effect, but only with the warning, “For use by domain experts or expert data analysts only; may cause undue distress or foolish optimism.”


What do you think?

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.