Building a Successful Predictive Analytics Program

Print Friendly, PDF & Email

JaneHendricks_IBMIn this special guest feature, Jane Hendricks, WW Portfolio Marketing Lead at IBM Predictive Analytics, describes a methodology for realizing business value from predictive analytics: start by understanding the business before the data that’s available and obtainable, then develop and apply models while considering how the models can be put into practice. The article includes three short case studies that illustrate the successful application of these principles. Jane is the WW Portfolio Marketing Manager for IBM SPSS Predictive Analytics which puts advanced analytics capabilities spanning ad-hoc statistical analysis, predictive modeling, data mining, text analytics, entity analytics, optimization, real-time scoring, machine learning and more into the hands of business users, data scientists, and organizations. This powerful foundation is the key to infusing predictive intelligence into everyday business decisions made by people and systems. Jane has over 20 of experience in advanced analytics spanning a variety of roles. She holds a Bachelor of Arts degree from Northwestern University in Economics and Political Science as well as an M.B.A. from DePaul University (Chicago) in Marketing.

Big data and predictive analytics are all the rage. With so many technologies and methodologies available, how can a business know that they’re making the right choices?

According to a recent Gartner survey, more than three-quarters of companies are investing or plan to invest in big data over the next two years. Respondents further indicated that they are pursuing multiple goals such as enhancing the customer experience, streamlining existing processes and reducing costs.

To realize the value of predictive analytics, it is critical to steer clear of the hype and focus on best practices in big data and analytics that will work for your business. While predictive analytics brings with it a rich arsenal of techniques such as machine learning, simulation, and clustering, they are merely a means to an end. That end is realizing business value through increasing the understanding of business processes, products, services, and customers in order to improve business efficiency, enhance customer service and ultimately, increase profit.

Here are the necessary steps to creating a successful predictive analytics program:

Start with a Business Problem

Business leaders must take a “bottom up” approach. Rather than applying analytics to a random data set without a specific end goal in mind, it is critical to start with a business problem and use analytics to solve it. Identifying priorities such as streamlining operational processes or understanding customer behavior will help refine the optimal end result. The more specific the inquiry, the more clear the answers will be.

When the Tennessee Highway Patrol (THP) turned to predictive analytics to improve highway safety and decrease accident rates without increasing staff, it started by asking very specific questions. To identify the best times and locations to deploy troopers, THP built a predictive model for traffic accidents that looked specifically at geotagged data on historical crashes and DUIs. The model identified problem hotspots, helping the THP more accurately deploy the DUI and seatbelt enforcement programs, helping the THP drop traffic fatality levels in Tennessee to the lowest level since 1963 and increase DUI arrests and safety belt citations by 34 percent.

Discover Data Sources

Once the business challenge has been identified, it’s time to focus on data sources. Both structured and unstructured data can hold great value, so it is important to not overlook external public data that can be combined with internal sources to yield greater insight.

Cable network AMC took this approach in an effort to develop higher-quality content by seeking new data sources to assess viewer habits. Instead of relying on traditional ratings data and third-party analytics, the company obtained raw data from Nielsen and ComScore and combined it with internal data from its live streaming and video on-demand services. AMC also merged the data with additional third-party data from partners such as iTunes and Amazon and streaming services like Neflix and Hulu. The company was able to gain deep insight into the viewers of each show. AMC is now able to better predict which shows will be successful, determine optimal scheduling and execute a more targeted marketing strategy.

Operationalize Insights

The last step is to develop and apply analytics models while considering how they can be operationalized. Successful analytics programs are built to turn insight into action.

Tesco Ireland, part of the larger Tesco PLC, which owns 2,900 supermarkets and retail shops, did this when they discovered large variations in the way refrigerator temperatures were managed across stores. The company harvested data from one store and applied predictive analytics to the data set in order to validate refrigeration performance and identify anomalies. Tesco was then able to enforce standard policies and control strategies on those anomalies. Ensuring that the freezers operate at the right temperature helped the company cut refrigeration energy costs by up to 20 percent.

The Key to Success

Successful analytics programs provide measurable results with real business value. Success is only achieved by combining an understanding of business objectives with available data to yield insights that can be turned into action. Measuring success is quite simple: if the program derives business value, it can be considered worth it.

 

Sign up for the free insideBIGDATA newsletter.

Speak Your Mind

*