Experts and technology professionals at the recent Marketing Analytics 2010 conference in Boston were overflowing...
with advice on how to design and execute an effective predictive analytics strategy.
Conference speakers described the importance of identifying business challenges and reporting results; maintaining a strong focus on data; following a basic process; and taking a methodic approach to choosing projects. One speaker also offered a preview of new predictive analytics software that will help organizations make sense of social networks.
A common theme throughout the event was the idea that enterprises with a bias toward action are most likely to have success with predictive analytics. In other words, companies that use analytics tools to come up with a reasonable forecast must have the nerve to act on that information. But this can be more difficult than it sounds. Conference speakers said that getting a typical business user to act on predictive analytics information requires training, patience and resolve.
SearchBusinessAnalytics.com was on hand at Marketing Analytics 2010 in an effort to distill the information into handy tips that can help with everything from planning to executing a predictive analytics strategy. Just remember, organizations with the “resolve to act” will probably get the most mileage out of the top five predictive analytics strategy tips.
Identify business challenges and report results
Anyone planning a predictive analytics program should begin by identifying the business challenges that advanced forecasting technology will address, according to conference speaker Jean-Paul Isson, the vice president of global business intelligence (BI) and predictive analytics at Monster Worldwide Inc., a popular online career assistance and recruiting firm.
For more on creating a predictive analytics strategy
Isson, who designed Monster’s global analytics program, said that an understanding of the business challenges being addressed will guide every subsequent decision on the path to predictive analytics, including technology and personnel decisions. A company might want to use predictive analytics to identify future sales prospects, achieve a greater understanding of customer behavior or figure out how many widgets to buy next year. Whatever the reason, Isson said it’s important to move forward gradually, taking on one business challenge at a time.
IT personnel and business users should collaborate at all steps of the planning process, from identifying business challenges to designing reports, Isson said. IT departments should also plan to regularly produce the results of predictive analytics incrementally. Organizations that take on overly ambitious projects risk losing traction and focus, he said.
“Every three months, at least, we have to show something to the business,” Isson said. “The business can’t wait.”
Maintain a strong focus on data
Finding “the right data” requires data mining and mapping out any pertinent internal and external data sources, while always keeping the relevant business challenges in mind. Business and IT cooperation is essential at this juncture, Isson said, and organizations should form a cross-functional team of IT and business staff to execute the process.
“The data is the foundation,” Isson said. “When you’re building a house, if your foundation is not solid, the house will collapse.”
Follow a basic process when executing the analytics strategy
Alberto Roldan remembers a time about 10 to 15 years ago when a rash of fledgling BI projects were failing at a rapid pace. Roldan -- the head of North America enterprise analytics at Teaneck, N.J.-based Cognizant, a large technology and consulting firm -- said that the reason those organizations failed at BI was because they didn’t remember to follow a basic process.
The lesson, Roldan told conference attendees, is that organizations interested in launching a predictive analytics strategy should remember to follow the basics or they will also be doomed to failure.
“There is a process to doing [predictive analytics] and the process is a fairly straight type of thing,” Roldan said. “You need to have an entry point, you need to have data governance, you need to have meta data, you need to have change requests, you need to have proper documentation and a way to capture those things.”
Take a methodic approach to choosing predictive analytics projects
Charlie Veers, a manager in the customer transformation unit at New York-based Deloitte Consulting Ltd., agreed that it’s important to take on small predictive analytics projects one at a time. He also had some advice on how to pick the right analytics project to move forward.
According to Veers, it’s a good idea to build a business case for each potential analytics project, then compare them with each other and pick the one that offers the best value.
“There is always a qualitative component and a quantitative component where you look at what are the benefit drivers for what I think I’m going to achieve, and then what are the costs involved,” Veers explained. “By using something quantitative [...] you’re able to look at those [projects] on a level platform and decide which one is going to benefit your company the most.”
Get ready for social network-focused predictive analytics software
Several text analytics software companies offer ways to conduct sentiment analysis of the unstructured data found within Facebook posts and on other social networking sites. But according to Cognizant’s Roldan, there are currently no software companies with the ability to turn social networking data into a useful numerical variable that can be plugged into predictive models. Well, not yet, anyway.
“We’re working on it,” Roldan said, “and we’re working on it in a collaborative way. [...] I think [a solution] is going to come out of signal detection -- how to transform signal detection into a numerical value. I think we’re going to be doing the same thing with social media. Give us about a year and I think we’ll be there.”