The Now and Next for Predictive Analytics
As data collection grows easier, predictive models become more viable. However, there's still work to be done.
Digital technology—and general Internet culture—continues to progress in waves. With no crest in sight, marketers continually perfect their strategies or risk getting swept up in torrents of data. One of marketers' most effective tools: predictive analytics.
“Predictive analytics is about the consumer, and bringing a holistic view of that consumer. To do that you have to understand the profile of the consumer. They are more than a segment,” explains Emad Georgy, SVP of product development and global head of development at Experian Marketing Services. “Technology has caught up with that concept. Now with big data, we have technology that's bringing true predictive analytics to the table.”
Predictive analytics isn't a new concept, but the exponential growth of technology has infused it with even more power. With the right data, marketers can gain an incredible edge. Even better, they can capitalize on predictive analytics to deliver truly customer-centric marketing. But first, marketers must confront predictive analytics' sobering realities.
Increased tech capabilities (should) enhance customer experiences
Marketers need to understand their data and, more importantly, how this data contributes to better customer experiences. The big data craze helped push today's trends of personalization and relevance; but marketers need not forget data and foundational lessons from the past.
“[Marketers can use] historical data to determine trends and patterns in a campaign's consumer base that can unveil valuable insights,” says Marcelo Wiermann, chief technology officer at mobile ad platform The Mobile Majority. “Another possibility is to use historical data [and] machine learning to preemptively identify where, when, and how an ad campaign should run.”
Of course, the customer should remain at the center of this conversation and the focus on exceptional customer experiences. “You have a lot of folks talking about channels and big data, and we've kind of lost our way. We're not thinking about the customer,” Experian's Georgy notes. “Marketing optimization has been the focus [of predictive analytics]—and that will stay relevant—but there's a bigger play; understanding the consumer. What is the best performing customer segment? Who are they? Is this group trending toward to becoming that? These answers can be transient.”
Predictive analytics isn't for everyone
The amount of data, varying budgets, campaign scale, and company size can play huge roles in whether predictive analytics will prove beneficial. “Predictive analytics relies on having a good, healthy amount of rich data beforehand to extract learning from. So, if the data isn't in place first, then predictive analytics has nothing to base its process on,” Wiermann explains.
Some small businesses may not have the resources to extract enough customer insights from big data, and other roadblocks include small target audiences or lack of commitment to a predictive model. Even an overly simple data collection solution could prevent predictive models from working, says Stephen Yu, associate principal at knowledge process management company eClerx.
“Depending on the industry, the prospect universe and customer base are sometimes very small in size, [or] the budget is just too tight to effectively employ a predictive model. [But], if data is inadequate for advanced analytics, there may be a major failure in data collection, or the data is too unstructured to yield meaningful answers,” Yu says.
Room for improvement
Although viable and promising, predictive models do stand to improve. Much of this improvement will continue to come simply by virtue of the changing times.
“Speed to implementation is the key challenge of predictive analytics. Implementation speed is a matter of processes, and surrounding elements like data hygiene, standardization, categorization, tagging, data summarization, variable creation,” Yu explains. “Accuracy of predictions must be improved. Accuracy requires more than modeling techniques. The quality of data matters far more than differences between modeling techniques. In predictive analytics, there is a price for being wrong, and the price can be severe.”
“Predictive analytics starts and ends with data, and I think the industry understands that,” Georgy says on a final note. “Data has scale, but coming back to the fundamentals of understanding customers, make sure your data supports that.”