Cutting Through Social Media Noise

Companies that adjust their products or services based on social media feedback might as well be driving blind. Award-winning research from marketing professor Wendy W. Moe shows the correlation between unadjusted social media data and well-developed offline market research is almost zero (0.008 to be exact). 

Despite the poor correlation, Moe said companies can clean the clutter that dominates the Internet and extract fast, affordable and accurate consumer data when they control for basic human tendencies that shape online conversations. 

The key is to apply the same rigor of traditional social science, refined over a period of decades, in the relatively young realm of Twitter, Amazon, Yelp and other virtual gathering spots. When market analysts do this, Moe said the alignment with traditional data sources improves one hundredfold. 

“If we ignore the effects of human behavior and just take social media comments at face value, we will get mostly noise,” said Moe, director of the Master of Science in Marketing Analytics program at Smith. “After we control for those factors, what we have left over is a brand measure that tracks closely with well-developed offline surveys.” 

Moe and her co-authors already have gained recognition for their research with “Social Media Intelligence: Measuring Brand Sentiment from Online Conversations,” winner of the 2014 Robert D. Buzzell Marketing Science Institute Best Paper Award. 

They also share their findings in a new book, “Social Media Intelligence” (Cambridge University Press, 2014), which shows companies how to control for three main human effects. 

Extremity effect: Extremely negative people go online to talk, but so do extremely positive people. The majority of consumers, the ones in the middle, usually keep quiet. 

Venue effect: When people interact, they tend to say things that they think their audience wants to hear.Over time, they gravitate to venues that support their opinions.Often the result is an echo chamber, where communities reinforce each other’s views and grow increasingly confident in their extreme positions.Savvy social media analysts control for the venue effect by monitoring a range of discussions in multiple locations. 

Topic effect: Public sentiment is more complex than positive, negative or neutral. People can express extreme views on a particular topic but still have different feelings toward the overall brand. Online market researchers must learn to distinguish the parts from the whole when analyzing brand sentiment. 

When companies control for these behaviors, they can convert the Internet into a rich source of continuous intelligence. “Companies can clean the data,” Moe said. “They can move from social media monitoring to social media intelligence.”