Pitfalls of predictive analytics

Remember the days when a rear-view mirror was all we needed to make business decisions? Now, predictive analytics appears poised to turn hindsight into a relic of the past.

Two Gartner analysts echo that sentiment, stating, “Few technology areas will have greater potential to improve the financial performance and position of a commercial global enterprise than predictive analytics.”

Executives are eager to jump on the bandwagon too. Although only 13% of 250 executives surveyed by Accenture said they use big data primarily for predictive purposes, as many as 88% indicated big data analytics is a top priority for their company. With an increasing number of companies learning to master the precursors to developing predictive models — namely, connecting, monitoring, and analyzing — we can safely assume the art of gleaning business intelligence from foresight will continue to grow.

Amid the promises of predictive analytics, however, we also find a number of pitfalls. Some experts caution there are situations when predictive analytics techniques can prove inadequate, if not useless.

Let’s consider three examples:

1. Predictive analytics works well in a stable environment in which the future of the business is likely to resemble its past and present. But Harvard Business School professor Clayton Christensen points out that in the event of a major disruption the past will do a poor job of foreshadowing future events. As an example, he cites the advent of PCs and commodity servers, arguing computer vendors who specialized in minicomputers in the 1980s couldn’t possibly have predicted their sales impact, since they were innovations and there was no data to analyze.
2. Bias in favor of a positive result is another danger when interpreting data; One of the most common errors in predictive analytics projects. Speaking at the 2014 Predictive Analytics World conference in Boston, John Elder, president of consulting firm Elder Research, Inc., made a good point when he noted that people “‘often look for data to justify our decisions, when it should be the other way around.”
3. Mining big data will further do little good if the insights are not directly tied to an operational process. I’ve a feeling more companies than we realize are wasting precious time and manpower on big data projects that are not adequately understood, producing trivia rather than actionable business intelligence.

With the above challenges in mind, talent acquisition and thorough A/B tests will be key components of any predictive analytics project. What else do you think organizations need to do to use foresight effectively?

You may also be interested in:

• Avoid the 3 Most Common Snags of Predictive Analytics
• Data Overload: The Challenge of Managing Supply Chain Analytics
• Using big data to solve problems
• How the Internet of Things will change your world
• Internet of Things and its impact on supply chain management