Predictive Analytics: Staying Ahead of Your Customers

Page 1 of 2

Article ImageCan a relatively mature technology help content publishers and marketers make website visitors more sticky and allow them to retain digital subscribers while also raising prices? The answer is yes.

The science behind what is making the aforementioned possible--predictive analytics--has been around for quite awhile. In its former life, it was known as data mining. Add in Big Data and the rapidly maturing technology looks as if it's ready for its close-up.


Predictive analytics (PA) "describes a range of analytical and statistical techniques used for developing models that may be used to predict future events or behaviors," as defined by Techopedia. It's been successfully employed in many industries, including banking, telecomm, and healthcare.

Data scientists create computer models to intelligently anticipate what a website visitor will want to see or read next. Here's an example: A visitor looked at an article on fashion the first time she visited a site. Thanks to PA and Big Data-using web analytics, social media, comments, and customer relationship management (CRM) data-on her next visit, she will be recognized (whether she has authenticated or not) and served up more fashion content. She will also see other, seemingly unrelated content that people who fit her profile (location, reading history, and demographic info, etc.) have enjoyed. If the predictive model is well-constructed and the content is good, perhaps she'll turn into a fan and frequent visitor of the website.

Direct mail and credit card issuers have been huge proponents and users of PA for years, while experts say that digital content providers have been somewhat more reluctant to embrace the technology. "Capital One basically built itself on analytics and getting an edge over the competition from doing a better job of choosing who would be better customers," says John Elder, Ph.D., founder and CEO of Elder Research, Inc., a data mining and predictive analytics firm. "Last time I was there, they had 200 masters and Ph.D.s working on the credit scoring problem."


PA works well in predicting what content to serve up, experts say. PA algorithms can even make educated guesses about what a new site visitor will want to see.

"The more info that we have, the more precise we are," says Bob Dutcher, VP of marketing at Apigee, an API and predictive analytics provider. "We tend to have a fairly high accuracy rate on content to serve up to individuals. And we've found in many places our ability to accurately predict what people are interested in can go up 20% to 100% percent on top of what we're already getting based on adding in richer data as well, including social and third-party data."

New visitors, unsurprisingly, are more of a challenge, according to Dutcher. He says that Apigee's software creates what's called a "lookalike," taking any information on the user to put them into a category of similar site visitors. "Basically, what you say is that people who look like this individual based on the information that we have tend to like X, Y, and Z," Dutcher says. "And that's of course what you'd serve up. Depending on their interaction with that, you can quickly modify it to kind of focus in on what their true interests likely are. The first click helps a lot. Every single click tells you more about that individual."

PA can also help marketers deal with a common problem: email list fatigue. Knowing who's likely to unsubscribe is a powerful weapon. "So if you can predict who's going to unsubscribe, you can email them less often," says Eric Siegel, founder of Predictive Analytics World and author of Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die. "Whereas, the other people you might feel free to email more often."

Page 1 of 2