This week we held another webinar in our Best Thinker webinar series, this time on the topic of Predictive Analytics. The webinar included speakers such as Eric Siegel, the founder of Predictive Analytics World and author of Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie or Die; Kaiser Fung, the Vice President of Business Intelligence and Analytics for Vimeo and the author of the new book Numbersense: How to Use Big Data to Your Advantage; and Mike Liddell, the General Manager of Digital for NGP VAN.
As the founder of Predictive Analytics World and Text Analytics World, Eric is obviously well versed in some great case studies that leveraged social media data and predictive analytics. One example that the audience really enjoyed included some data around a US telecom company that uses predictive analytics to diagnose that a person is 700% more likely to cancel a contract with this telecom carrier if someone in their network also cancels their contract. It was dubbed “churn model performance with social data”!
Kaiser Fung from Vimeo went on to talk about how good outcomes require good data – which was a big hit with the audience. He discussed some of the pitfalls of certain types of data; for example, with click data you can’t assume that the click means that the person has expressed interest – perhaps they were on their mobile and clicked by accident (dubbed “the Fat Finger effect”) or perhaps they were tricked into clicking; you can also expect a percentage of clicks to be “fraudulent.”
Mike Liddell, the General Manager of Digital for NGP VAN, talked about the use of predictive social data in the Obama re-election campaign, and how social data and analytics played a big part in the most recent election.
If you have been on a SocialMediaToday webinar before, you know they are very “participant-driven” and we love your questions. Many of the questions from our audience revolved around topics like: Sentiment – have we really cracked the code? How do you ensure you have the best possible data / is there a way to verify data? The audience also wanted to know: Can you merge the influx of nontraditional data sources such as blogs and social networks with older legacy systems without sacrificing analytic integrity?