You've won approval for a predictive analytics program, and away you go on the implementation, fully expecting it to generate valuable business insights that lead to increased revenue and competitive advantages. But a year later, corporate management doesn't think the investment has panned out in the form of tangible business benefits. What went wrong?
In many cases, a rush to deploy predictive analytics tools without proper planning sets the stage for unmet expectations, according to technology analysts. They warn that not setting specific goals and fully measuring the business pulse of predictive analytics programs up front can crimp their potential before deployments even begin.
Thus, the initial challenge for an analytics team is establishing a set of strategies and processes that will logically advance a project toward a successful outcome. That means diligently thrashing out and defining both short- and long-term objectives with corporate and business executives to make sure the analytics efforts aren't headed down the wrong path.
Predictive modeling success must be earned
There's also the need for a sustained effort to develop, test and refine predictive models to ensure they generate findings in line with a company's business goals and requirements. "Organizations run into trouble when they expect to [automatically] get amazing results," said Doug Laney, an analyst at Gartner Inc. in Stamford, Conn. "Predictive analytics projects are iterative and involve processes that need the regular testing of models."
Typically, there is an assortment of business metrics and variables to take into account. Understanding the value and significance of the potpourri of available elements is no easy task, said Eric King, president of consultancy The Modeling Agency LLC. In fact, he added, it's tricky enough that a common problem is developing predictive models that answer the wrong questions.
For example, a model might tell a car maker that males younger than 23 years old prefer red interiors over turquoise ones. That might be an accurate result, but it won't be useful from a business standpoint if a substantial majority of prospective car buyers aren't predisposed to surrounding themselves with either of those colors. King said obtaining irrelevant information of that sort from predictive models is a waste of time, effort and resources that can be avoided only through close interactions between an analytics team and business managers.
Take it from the top on predictive analytics programs
Because the process of implementing and sustaining a predictive analytics program is inherently complex, a lack of executive support and leadership can hamstring efforts. What's also needed from the top ranks is the endorsement of a corporate culture that values creative thinking, fresh ideas and data-based decision making, said John Lucker, head of the advanced analytics and modeling practice at Deloitte Consulting LLP. "In such a culture, predictive analytics projects, and a lot of other positive things, flourish."
On the other hand, an unwillingness on the part of business executives to trust or credit predictive analytics findings can have a big impact on the bottom line. Lucker recalls facing resistance from a supermarket chain to a study he produced showing that customers would pay more for chocolate ice cream than vanilla. Executives at the company contended that varied pricing would add to its labor costs and be viewed with disfavor by shoppers. In the end, Lucker's recommendation was rejected -- leaving a few million dollars in potential revenue sitting on the table, he said.
Overcoming such resistance -- to change and to taking action based on where the data leads -- is a must for organizations determined to maximize the business value of their predictive analytics programs, according to Lucker. As a result, that needs to be a top priority of program managers. "We're imperfect human beings with our egos and so forth," he said. "But significant progress, in terms of better efficiency and smarter decisions, will come to fruition if we simply insist on making decisions based on facts."
Roger du Mars is a freelance writer based in Redmond, Wash. He has written for publications such as Time, USA Today and The Boston Globe, and he previously was the Seoul, South Korea, bureau chief of Asiaweek and the South China Morning Post.
Read up on the importance of sufficient training for predictive analytics success
Learn the recipe for success with predictive analytics
Are predictive analytics and event processing the future of BPM?