Alternative-data firms are shedding new light on corporate performance

Tip: this article is not in the category of computer technology, but I like.
提示:本文不属于计算机技术范畴，但我个人喜欢它。
FINANCIAL statements are both infrequent and backwards-looking, so getting a sense for how a business is performing in the present can be nearly impossible. But a cottage industry of a few dozen firms, mostly in America, is gleaning “alternative” data from novel sources, ranging from satellite images to obscure corners of social media.

The growth of small, low-cost satellites and machine learning means companies can quickly and cheaply parse millions of satellite images a day. A common trick is to analyse photos of car parks outside big-box retailers such as Walmart to get a sense of daily revenues. A Chicago-based data firm, RS Metrics, sells estimates on the productivity of factories by tracking the number of lorries parked outside. Bad weather can make such analysis difficult in some places, but satellite-image analysis of, say, Elon Musk’s new “gigafactory” making batteries for Tesla’s electric cars in Nevada is more straightforward under the desert’s clear skies.

A bit of ingenuity along with some elementary geometry goes a long way. Data analysts estimate the size of oil stocks by looking at the lengths of shadows cast by oil tanks in satellite pictures (the height of the roofs of most crude-oil tanks varies depending on how full the tank is). Several firms, such as Orbital Insight, in Palo Alto, also study farmland to estimate crop yields before official statistics are reported by America’s Department of Agriculture, and often do it better (see chart). Investors are particularly keen for firms to study pictures that yield rare data on, say, steel production in China or Russia, where official data can be patchy.

Dataminr, a startup in New York, mines social media for happenings on which to alert its clients, which include hedge-fund traders and big newsrooms. Twitter has taken a 5% stake in the firm. Early this year a local reporter tweeted that the FBI was raiding the offices of United Development Funding, a sponsor of real-estate investment trusts in Grapevine, Texas. It took other investors around ten minutes to hear the news and to push its shares down by 50%, by which time Dataminr’s clients had been able to short them.

Some social-media firms are themselves branching out into alternative data. Foursquare, which is known to consumers for its mobile app that provides restaurant recommendations based on its users’ locations and histories, now sells data. Foursquare can accurately guess if someone is a patron of a particular shop based on how long he has stopped moving (five minutes or more is the trigger).

Alternative-data firms also offer insights into private companies, such as technology “unicorns” (firms that have yet to come to the stockmarket but are valued at $1 billion-plus). Second Measure, based in San Francisco, claims it can show how many subscribers Netflix had this month, or how Uber, a ride-sharing service, is doing relative to Lyft, a rival. The information comes from data that Second Measure collects on credit-card transactions. For venture capitalists, alternative-data firms may be the only objective source of sales data.

There have been plenty of acknowledged triumphs. In a blog post earlier this year, the boss of Foursquare, Jeff Glueck, used his company’s foot-traffic data to predict, correctly, that same-store sales at Chipotle, a restaurant chain affected by an outbreak of E. coli, would fall by 30%. Shares in Chipotle fell by 6% when the company reported earnings.

One impediment to broad adoption of alternative data is a cultural divide between west-coast techies and buttoned-up east-coast financiers, notes the boss of one data provider. He was dismayed to find, on a visit to one richly-resourced fund, that it was guilty of what techies consider the ultimate sin: using Windows computers. Such data is also expensive, and the payoff can take time. But the value of the information to hedge funds and other investors is growing fast.