There is a ton of data out there about what you’ve bought, where you’ve traveled, the identities of your friends, and what you like to do. Thanks to new data-analytics tools, all this information can be used to predict what you, as an individual, will do next.
Advertisers and marketers value consumer data for insights on which advertisements to serve up, how to position products, and to better understand how consumers react to a brand. Thanks to the availability of better data-analytics tools and a growing interest in Big Data deployments, organizations are beginning to realize they have a goldmine of information at their disposal, Eric Siegel, former Columbia University professor and author of “Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die,” said in a recent interview.
The excitement around Big Data is “its potential to predict individual behavior,” Siegel said. “The ‘big’ in Big Data refers to its untapped potential and not just the volume of the information available.”
In predictive analytics, data analysts combine various sources of information to predict future probabilities and trends. With those predictions in hand, organizations can make operational decisions to improve operations and customer service.
For example, an insurance company can look at age, gender, and driving record of the driver to decide appropriate rates and policies. Retailers such as Target can look at shopping patterns to predict who may be pregnant, and connect them to baby-related offers. Or a major health insurance company may predict the likelihood of an elderly policy-holder dying within 18 months in order to trigger end-of-life counseling.
In a similar vein, the IRS uses predictive analysis to rank tax returns on the likelihood of fraud, allowing it to find more instances of tax evasion without increasing the number of open investigations.
Unlike forecasting, which ferrets out trends and patterns over a larger population, predictions are “specific to the individual customer, citizen, law enforcement suspect, or patient,” Siegel said. They also aren’t limited to just a handful of factors, such as demographics or shopping history, but consider “any and all aspects of the individual,” such as information posted on social media, public records, and personal behavior.
While some data pieces may be considered “mild evidence,” they form a much clearer picture when combined with other pieces. “Anything and everything has potential to contribute as another piece of evidence,” Siegel said.
The benefits of the right prediction tools to an organization are obvious. They can calculate risk more accurately, market to customers better, and allocate resources effectively. Data gives them power, Siegel said. And consumers receive immediate benefits, such as getting relevant offers, better healthcare, and improved spam and Web filters.
But there are also some negative privacy and civil liberties concerns, especially considering the sensitivity of the information—data needs to be secured. Individuals may be okay with certain pieces of information being available to their healthcare provider or insurance company, but may not appreciate hearing from a marketing firm. It may be a privacy violation if your manager and employer can make predictions based on health-related information.
Tech giant Hewlett-Packard collects information from various sources to predict which employees may be planning to quit the company and provide that information to the manager. “Usually, the manager is the last person you want knowing that you are on your way out,” Siegel pointed out.
Police departments in Chicago, Los Angeles, and Memphis use predictive analytics to determine which neighborhoods need to be patrolled, Siegel added. Prison systems and judges in Oregon and Pennsylvania consult predictive models to decide who goes to jail and how long the sentence should be. Using a mathematical model to determine when a person should be released from prison may infringe on civil liberties.
The issues at hand are not technological but sociological, Siegel said. There needs to be a frank and thorough discussion on how this technology can and should be used, and decisions made on how to preserve privacy and civil liberties. While predictive analytics can be used improperly, bottom line, “we are never going to outlaw knives, but knives can be used for evil,” he noted.
Regulation—how much and who should be in charge—will be necessary, but no one really knows what the best way is, or how to accomplish it. “That is something we don’t have the data to predict,” Siegel added.