Companies Derive Credit Scores from Phone MetadataPrompt bill payers are typically more reliable than those who hold off, and people who make frequent calls far outside a bank’s network are more likely to have trouble making deposits. However, even the esoteric information can factor in — researchers at Cignifi, a Cambridge-based firm studying the predictive capabilities of mobile data on loan repayment and savings, found that the time of day and neighborhoods from which calls are placed can be indicators, too.
Analytics moves from last touch to holisticThey were doing that analysis for some time actually with a method called “last touch.” That means identifying the last thing that the customer did before they bought—as in they clicked an ad and then they bought whatever. The company figured it must’ve been that ad that caused the customer to buy the print. Or someone got a direct mail campaign message and then they bought the calendar. That was the motivation. [Shutterfly] looked at that process and said, “You know, that’s a good model. It’s a good approximation, but it would be better to look at everything touching the user before their last purchase and since the purchase before that.” This greatly expanded the data that they had to consider to do the analysis, so the process became very slow. It took two days to compute the likely marketing channels for all their orders.
In fact, mapping attention to power in an organization gives a clear indication of hierarchy: The longer it takes Person A to respond to Person B, the more relative power Person A has. Map response times across an entire organization, and you’ll get a remarkably accurate chart of social standing. The boss leaves e-mails unanswered for hours; those lower down respond within minutes. This is so predictable that an algorithm for it—called automated social hierarchy detection—has been developed at Columbia University. Intelligence agencies reportedly are applying the algorithm to suspected terrorist gangs to piece together chains of influence and identify central figures.
Over the past several years, Purdue University has been experimenting with a data-driven solution way to find kids who are at risk for dropping out, or who--in a critical mass--might indicate which classes or majors have inadequate instructors. Administrators call it a “student success algorithm,” but it’s official name is Course Signals--and if it works, it could change the way modern universities are run. Incorporating data-mining and analysis tools, Course Signals not only predicts how well students are likely to do in a particular class, but can also detect early warning signals for those who are struggling, enabling an intervention before problems reach a critical point.
Topsy makes its money from more sophisticated tools — aimed at marketers, media companies, political operations, and hedge funds — that require a subscription fee that starts at $12,000 a year. Those allow searches that compare different terms, narrow down results by geography and surface the specific tweets with the most influence on the social conversation.