The role prediction played in the 2012 presidential election taught both the political and business worlds an important lesson: true power comes from influencing the future rather than merely predicting it.
The role prediction played in the 2012 presidential election taught both the political and business worlds an important lesson: true power comes from influencing the future rather than merely predicting it.
A crystal Ball with carved ivory dragon stand in the Anderson House on May 14, 2012 in Washington DC (Photo by Benjamin C. Tankersley/For The Washington Post)
More Read
Blogger Nate Silver may have successfully forecast the election results, but President Obama’s team quietly used predictive analytics to sway which way the winds would blow, discovering which individual voters were more likely to be positively influenced by campaign contact.
But this is only one example of how the prediction of individuals’ wants, needs and behaviors holds the power to change outcomes.Businesses, much like political campaigns, benefit from moving beyond the forecasting of broad trends to the forming of individual, per-person predictions. In the business world, these predictions drive the detailed operations of marketing, risk management, and fraud detection one customer at a time.
Now, rest assured, a doctorate isn’t required to understand how computers churn out tens of thousands—or even millions—of predictions. The principles behind predictive analytics are relatively easy to understand.
That said, the underlying math can get complex quickly. Analytics experts often favor a range of mathematical formulas to achieve desired results. Hewlett-Packard, for example, uses complex equations in order to predict the odds that each HP employee will leave his or her job. Credit scoring uses similar equations, predicting the probability of delinquent payments for each credit card or loan applicant. Meanwhile, mass marketing uses advanced analytical methods to target those customers most likely to buy. Common mathematical methods for these purposes include logistic regression and neural networks.
While IBM’s “Jeopardy!”-playing Watson computer does not predict the future, it does employ the same sort of analytical models to “predict” the correct answer to each question, learning from large amounts of data to infer unknowns. To supercharge its models, Watson “ensembles” them together, creating what could be called an “uber-model.” That uber-model benefits from the wisdom of a crowd of models in much the same way a crowd of collaborating people often succeeds to collectively surpass the intelligence of each individual member.
Predictive modeling methods in use today vary, but all follow the same basic principle: Data is a representation of experience—a recording of things that happened—and thus provides examples from which to learn. Computers then learn from these examples to predict future behavior, which serves to drive and automate more effective decisions.
