Five Trends in Predictive Analytics
Predictive analytics, a technology that has been around for decades has gotten a lot of attention over the past few years, and for good reason. Companies understand that looking in the rear-view mirror is not enough to remain competitive in the current economy. Today, adoption of predictive analytics is increasing for a number of reasons including a better understanding of the value of the technology, the availability of compute power, and the expanding toolset to make it happen. In fact, in a recent TDWI survey at our Chicago World Conference earlier this month, more than 50% of the respondents said that they planned to use predictive analytics in their organization over the next three years. The techniques for predictive analytics are being used on both traditional data sets as well as on big data.
Here are five trends that I’m seeing in predictive analytics:
- Ease of use. Whereas in the past, statisticians used some sort of scripting language to build a predictive model, vendors are now making their software easier to use. This includes hiding the complexity of the model building process and the data preparation process via the user interface. This is not a new trend but it is worth mentioning because it opens up predictive analytics to a wider audience such as marketing. For example, vendors such as Pitney Bowes, Pegasystems, and KXEN provide solutions targeted to marketing professionals with ease of use as a primary feature. The caveat here, of course, is that marketers still need the skills and judgment to make sure the software is used properly.
- Text hits the mainstream in predictive analytics. The kind of data being used as part of the predictive analytics process continues to grow in scope. For example, some companies are routinely using text data to improve the lift of their predictive models because it helps provide the “why” behind the “what”. Predictive analytics providers such as IBM and SAS provide text analytics as part of their solution. Others, such as Angoss and Pegasystems have partnered with text analytics vendors (such as Lexalytics and Attensity) to integrate this functionality in their products.
- Geospatial data use is on the rise. Geospatial data is also becoming more popular for use in and with predictive analytics. For instance, geospatial predictive analytics is being used to predict crime and terrorism. On the business front, location based data is starting to be used in conjunction with predictive modeling to target specific offers to customers based on where they are (i.e. traveling from work, at home) and their behavior.
- Operationalizing the analytics for action. Operationalizing means making something part of a business process. For example, companies are using predictive analytics to predict maintenance failures, predict collections, predict churn, and the list goes on. In these examples, predictive models are actually incorporated into the business process of an organization. For example, if a customer takes a certain action that puts them at risk for churn, that customer’s information is routed to the appropriate department for action. In fact, the term “action” and “insight to action” has come up quite a bit in recent conversations I’ve had with vendors.
- Adaptive learning: I’ve heard this go by a number of names – adaptive intelligence, automated learning, and adaptive learning. The idea is about continuously learning. For example, a model to understand behavior might be deployed against customer data. As the data changes, the model might change too. This kind prediction could also work against streaming data. Adaptive intelligence is still pretty early in the adoption cycle, but I expect it to increase.
These are just a few of the trends that I’m seeing in predictive analytics. As the technology continues to be adopted, new trends will certainly emerge. I used predictive analytics back in the late 1980s when I was at AT&T to understand customer behavior and I’m very happy to see that it’s a technology whose time has finally come! I’m now starting work on TDWI’s Best Practices Report on Predictive Analytics. Expect more from me on this topic in the future.
Posted by Fern Halper, Ph.D. on May 22, 2013