Foursquare Shows That ‘Alternative’ Data Is Now Mainstream

The accurate prediction of Chipotle’s profit drop suggests a different future for data


Foursquare hit the news last year when it analyzed its foot traffic data alongside Apple’s public sales data to accurately predict the number of iPhone 6Ss that would sell on its launch weekend. It claimed that footfall around Apple stores around the world would more than quadruple from preceding weeks, translating into sales of between 13 million and 15 million. Wall Street analysts predicted that it would be around 12 million. The final number was 13 million.

It appears that Foursquare has managed to replicate the feat again, accurately predicting restaurant chain Chipotle’s drop in profits with an extraordinary degree of accuracy at 30% - just 0.3% off the actual figure of 29.7%. While one success may be easy to write off, they suggest a trend of abstract datasets being more accurate than traditional data prediction models.

While it was Chipotle’s first loss as a public company, the signs were there that it would be a difficult quarter from last year. Chipotle prides itself on its food being chemical free, and reports of E. coli among diners last winter were always going to have a negative impact on sales. A federal criminal investigation was also launched in relation to a norovirus outbreak in California. Despite this, investors were still caught unawares, and shares dropped 6% on the news. Foursquare were not.

Foursquare’s predictions work on a fairly simple level. It uses its own proprietary technology, called Pilgrim, to find a user’s location by looking at historical check-in data, eventually working out that a given location represents a certain store. Using this knowledge, it can tell when the smartphones of other Foursquare users go there even if they don’t check in by looking at GPS signal, WiFi, cell towers and beacons to pinpoint where smartphone users are. When more people visit a store, it would be easy to expect that, by looking at the data alongside other data sets, sales would go up. Equally, as in Chipotle’s case, it is easy to see when they are going to go down.

Since the company launched seven years ago, Foursquare has collected 85 million venues and 50 million monthly users in its system. It describes its data trove as the ‘biggest foot traffic panel in the world,’ a claim that’s hard to disagree with. In the Medium article in which Foursquare CEO Jeff Glueck made the Chipotle prediction, he was optimistic in his assertion that the data suggested the firm would recover, but they needed to lure back loyal visitors or nurture new fans. Whether they do this, it’s hard to predict. For the cottage industry of tech firms that has formed in recent years to exploit data produced by smartphones and the billions of other data sensors that now measure every aspect of the world, the future is far more obviously a bright one. These firms process information on everything from the weather to people’s shopping habits in-store, and can sell it for thousands of dollars. Increasingly, they will play the central role in Wall Street analysts’ models as they look to gain an advantage, but it’s not just larger firms that they could reveal insights about. On its own, the scale of the data needed is likely to make it difficult for Foursquare to produce revelations about smaller organizations than Chipotle. However, by looking at these data streams alongside one another, the scale of data and the points of contrast increase, making it easy to spot trends, and it’s highly possible that this could produce insights about smaller firms too. The potential is ultimately limitless, and Foursquare’s successes are probably only the tip of the iceberg.

University lecture small

Read next:

How Are Higher Education Institutions Using Analytics?