To do so, the consulting, market research and analyst firm selected 12 vendors it sees as strong predictive analytics contenders that could also provide a range of strategies and price points for potential buyers. The firm then measured the vendors’ vision, viability, validity and value across 50-plus attributes by talking with customers and executives, reviewing surveys and factoring in chatter from social media sites using a platform called NetBase. Although the decision to include some vendors may come as a surprise -- Megaputer Intelligence Inc., for example -- Hurwitz & Associates’ selection of predictive analytics for its first index report shouldn’t.
“There’s a great deal of business value in these advanced analytic-type of techniques -- predictive analytics being one of them,” said Fern Halper, a partner for the Needham, Mass.-based firm. “We think it’s a market that’s ripe for growth. It’s moving out of the early adopter stage, certainly, and into the mainstream.”
While the bulk of the Victory Research Index Report details the ins and outs of each vendor’s offerings, Halper, the principal researcher for the study, also documented trends and said the rising interest in predictive analytics is based on changes in both technology and culture.
“Especially in today’s economy, companies are realizing they can’t just look in the rearview mirror and look at what has happened,” said Halper. “They need to look at what can happen and what will happen and become as smart as they can possibly be if they’re going to compete.”
Business analysts, meet predictive
One major trend Halper noted is a shift in how businesses are shopping for predictive analytics tools. In the past, building predictive models required statisticians or quantitative analysts, but today, companies are also interested in putting those tools in the hands of their business analysts who may not have the same mathematical background.
Vendors are responding by incorporating sleeker user interfaces, found in products from SAS, SPSS and StatSoft Inc.; automating options that can size up the data and recommend the appropriate model or models to use for analysis; and providing cloud-based services to support businesses that don’t have the skills to analyze data in house, an option offered by Angoss Software Corp. for example.
“But the way you really get predictive analytics to be more pervasive in an organization is by embedding it in part of the business process,” Halper said. “So operationalizing the model.”
Embedding predictive models into the business process means providing actionable information on how to respond to a prediction. In the case of customer dissatisfaction, for example, a model may pinpoint a potential unhappy customer. By embedding the model into the business process, an employee may be alerted to offer that customer certain products or services to curtail churn.
Business analysts, meet the predictive models
Some companies not only want their business analysts to have access to predictive analytics tools, they want them to build the models as well.
“In many instances, the business analyst will build something, and, if they want it to be more accurate and if the company has a statistician in house, the statistician may be the one who refines the model,” Halper said.
While this method may help free a statistician from building every predictive model from scratch, Halper also cautions, “a model in the wrong hands could be a disaster.” If businesses go with this approach, she advises providing employees with the appropriate training.
“If there aren’t enough safeguards in place, it could be dangerous,” she said. “You need to know your business, you need to know your data and you need to know enough about the models you’re running to know if what you have makes sense in terms of output.”
This is only one technique for more fully inserting predictive analytics into an organization, Halper adds. Other models, such as the one provided by TIBCO Software Inc., enables statisticians to build models while providing business analysts with mash-ups, easy-to-use interfaces to visualize and interact with the data. These are prevalent and may be a good fit for an organization.
Big data, model management
The practice of predictive analytics combined with “big data” may be spurring businesses to turn to high performance computing, which, in recent years, has become more accessible, Halper said. This has helped to push advanced analytics into the mainstream.
“A number of companies would throw data at some of the vendor products, and those products would crash their machines,” she said. “They realized they were going to have to move to 64-bit machines and other high performance computing to do that big stuff.”
Now, Halper said, businesses that have deployed a predictive analytics strategy are searching for ways to manage the potentially thousands of models they’ve built, a demand vendors are again striving to meet.
Most software vendors offer methods of checking models into directories, but software vendors like SPSS and SAS are also helping organizations track the metadata of who created the model and when, and even alert end users when a model may become stale.
“Some vendors are further along with this than others,” she said. “Some have a whole factory approach to model building. If you’re going to be building models around certain marketing campaigns, the software will let you make sure those models are tightly controlled: You know where they are, you know what they are, what they’re doing and when they’re not useful anymore.”