August 22, 2014 | Written by: THINK Leaders
Categorized: Data | How To | Marketing
An idea that goes viral—spreading uncontrollably through social networks—can touch millions of people per hour. Some viral ideas are positive. But many are complaints about specific companies and have the potential to damage sales and brands.
The challenge is to identify potentially damaging ideas before they spread to a national or global conversation.
Emerging technologies will allow organizations to manage their social risk with a robust system that monitors social networks in real time, predicts which ideas are likely to go viral, and decodes the meaning of a potentially viral idea so that a company can take action before damage occurs.
Step 1. Isolate a topic and map the conversation
Viral prediction starts by isolating a particular topic—such as a brand name, ad slogan or particular event—and mapping the conversation. Understanding and even visualizing the shape of a particular dialogue is important because polarized, dispersed and unified conversations behave differently.
A recent study by Indiana University analyzed hashtags in tweets during quickly evolving conversations to best understand the shape of the dialogue. The ongoing back-and-forth about the #GOP, the U.S. Political Party, was polarized while information shared during the uprising in #Egypt was dispersed.
Step 2. Understand the social sentiment
Analyzing the sentiment of a conversation can produce a snapshot of the public’s opinion. Social sentiment tools analyze millions of public posts and use semantic clues to determine the percentage of a conversation that is negative and positive so you can understand how the public is reacting in real time and focus on potential damage or opportunity.
At the USC Annenberg Innovation Lab, Professor Jonathan Taplin analyzed the social media chatter about upcoming films like Footloose, The Three Musketeers and The Lion King 3D to predict if the public would like them on opening day in time to allow for changes in marketing.
Step 3. Identify the anomalies
Out-of-the-ordinary sharing activity often indicates the start of a viral event. The viral prediction system uses algorithms to rate and track each individual comment within a conversation. When a comment from an individual with a modest following sparks a disproportionate amount of sharing or an individual with a large following broadcasts a message, the system raises a red flag and triggers closer inspection.
Step 4. Understand the meaning
Emerging technologies, such as statistical text analytics and natural language processing, allow the system to understand the meaning of the flagged social comment and analyze the combination of words to predict its potential to spread.
Natural language processing can interpret common idioms and sarcasm—”crying tears of joy” can be interpreted positively even though “crying” is usually a negative reaction. It can also flag a comment for review when an unusual combination of words is detected. “I just bit into a scone and found staples baked into it” would be flagged since “scone” and “staples baked” are very unlikely to be together.
Step 5. Predict the timeline
It is becoming possible to forecast how long it will take for a conversation to spread by analyzing the size and influence of the network and comparing it to past viral events. This is an essential input for models that inform the next best action.
Step 6. Determine how to act
The final piece of viral prediction is to determine what to do about the imminent viral message. Emerging technology can provide early warnings and even prompt responses. Any action – continued monitoring of the dialogue, broadcasting an independent message or modifying a marketing campaign – should be informed by a complete picture of the viral event, the sentiment driving it and its trajectory in the social environment.