What is alternative data?

The word is relative to what investors traditionally read. The SEC's examination staff described it in a 2022 risk alert as many different types of information used in financial analysis beyond financial statements, company filings and press releases.

Their examples included satellite and drone imagery of crop fields and retail car parks, aggregated card transactions, social media and search data, phone geolocation, and email data from consumer apps.

A working definition

In practice it is any dataset that tells an analyst something about a company's revenue, costs or demand before, or in more detail than, the company reports it.

Examples and buyers

A freight operator's lane volumes, a booking platform's reservation counts or a recruiter's job-ad tallies can all qualify. The buyers are mostly hedge funds, asset managers, private equity firms and the research teams that serve them.

Investment firms worry about material non-public information. If a dataset contains confidential information that was obtained in breach of a duty, trading on it can raise insider trading issues.

What regulators criticised

The SEC risk alert of April 2022 criticised advisers for ad hoc diligence on alternative data providers and for not reviewing the terms under which data was collected. That is why serious buyers send long questionnaires before they sign.

Sellers face scrutiny too. In September 2021 the SEC settled what it called its first enforcement action charging an alternative data provider with securities fraud. It found that App Annie told app companies their data would be aggregated and anonymised, then used non-aggregated data to adjust the estimates it sold to trading firms.

The lesson for anyone supplying this market is that the promises made to the original data sources matter as much as the promises made to buyers.

This walk-through is fictional. A private chain of car washes operates sites across several states, and its point-of-sale system records every wash by site, package and hour. Because many of its sites sit beside big-box retailers and fuel stations, daily volumes track local traffic in a way an analyst covering those listed retailers might find useful.

The broker prepares the seller for diligence. The records are the chain's own, so ownership is clear. Card numbers and licence-plate captures are excluded; the product is counts and package mix per site per day.

The chain is private and has no confidential agreements with the nearby retailers, which reduces the risk that the data embeds someone else's non-public information. The broker writes up a sourcing memo describing how counts are generated, what changed when the point-of-sale system was upgraded, and which sites opened or closed.

An investment firm trials a historical sample, tests whether it lines up with what the retailers later reported, and only then licenses an ongoing feed.

The licence terms

The contract bars resale, names permitted users, and lets the firm audit the sourcing memo each year. Whether the signal is strong enough to pay for is the buyer's call; the broker's job was to make it trustworthy enough to evaluate.

• Whether the seller is listed, or holds confidential information about listed companies under contract.
• That individuals cannot be picked out, especially in location, card and email-derived panels.
• History length and consistency: investors want several years with breaks explained.
• What the original sources were told, in writing, about how their data would be used.
• That a sourcing memo and sample are ready before the first buyer call.