Follow Datanami:
May 15, 2012

Dispelling Predictive Analytics Myths

Datanami Staff

Even with the most comprehensive, precise, and data-laden models, predictive platforms are only able to present the best guess given a particular set of conditions.

Despite best hopes and efforts (on the part of marketing and technologist folks alike) predictive analytics software comes with no real guarantee of accuracy—and to date, there is little on the technology horizon to bring the accuracy to 100%.

Still, there have been some impressive feats of predictive modeling prowess in the last few years, tackling everything from consumer migration to political activity—all with stunning accuracy. But amidst all the attention paid to the power of predictive analytics as a concept, there are some critical talking points that are too rarely addressed, at least according to a few thought leaders in the business analytics space.

If you ask Piyanka Jain, the former business analytics lead at PayPal and current CEO at Aryng whether or not there is real business value in predictive analytics, her response to the body of predictive technologies will be enthusiastic—but with a few caveats.

Jain thinks that while the basic value of predictive analytics is easily recognized, the actual implementation of predictive analytics solutions is not as informed or well-guided as it should be. As she noted recently, “predictive analytics is often left to the devices of data miners and data scientists and hence is often misunderstood and misued by businesses.

All of this aside, she stresses that predictive analytics isn’t something that just emerged with the big data craze. She contends that predictive modeling techniques date back thousands of years (citing the use of Indian astrological charts in arranged marriages as a prime example of yore). Moving closer to the present, credit scoring models from the 1930s that leveraged predictive modeling techniques (as developed by Fischer and Durand) still look rather modern, minus the algorithmic technology component.

Jain says that “Good predictive analytics software tools do necessarily equate to good models.” She continued, “ With tremendous development in the software tools front with better graphical user interfaces (GUIs) as well as higher automation, new users often mistakenly believe that a good model can be built automatically by pressing the GUI “build model” button. But that is far from the truth. Building good models requires proper technical skills and use of a model building process. Though surprisingly, sometimes even that does not deliver a good enough model.”

Many of the myths that swirl around predictive analytics are even more dangerous now since more platforms are putting the power of predictive analytics in the laps of enterprise users who have never considered themselves data scientists. While this “democratization” of data can be incredibly beneficial, there are some important hardware and software considerations that need to be made on the part of IT in advance.

As Jamie MacLennan, CTO at Predixion Software says, there are some hard technical problems with the way that many enterprises are approaching the use of predictive analytics platforms. MacLennan says that first and foremost, many organizations feel that they need to have “a heavyweight data infrastructure in place” to apply predictive models—and “they spend years trying to build and cost justify the existence of a data warehouse, including the component applications, and then miss the mark entirely.” In this view, Modern predictive analytics tools don’t always need the perfect storm of a complete data infrastructure to provide meaningful and immediately applicable results. Today’s tools can handle a much wider variety of data situations. LacLennan says that many allow end users to glean insights into their own personal data while working with familiar tools — such as Microsoft Excel. With self-service, easy-to-use predictive analytics tools, any and all information workers can leverage predictive analytics with the day-to-day data they have at their disposal.

On that note, MacLennan says that as the democratization of data has continued and new tools have emerged, one of the biggest myths about predictive analytics—that it involves months and months (not to mention thousands of dollars in human and IT capital)—still survives. To counter this myth, he says…

“Predictive analytics has evolved from a back-room activity (where results were embedded into the deepest level of an organization’s business processes) to a tool that knowledge workers can use every day to help them make better decisions about actions to take right now. Although traditional, large-scale predictive projects still provide great value and competitive advantage, modern predictive analytics software allows you to deploy predictive analytics to all your end users, offering them a better understanding of how data drives your business and helping them make better choices every day.”