USA today ran an article Tuesday titled "Can social media predict election outcomes?"
The article predicted that on Super Tuesday Romney would be the top vote-getter in seven states and each of the other candidates would win one state:
North Dakota: Romney
Alaska: Ron Paul
These predictions were according to an analysis of Twitter by a social analytics software product. The research drew from more than 800,000 tweets from the past week. The social analysis measured positive sentiment and the share of voice for each candidate.
Were the predictions right?
Half the predictions were incorrect. A failing grade.
The obvious reason is that social media data does not yet provide a representative sample of the general population. But even if it did, making predictions based solely on the 140 characters in a bunch of Tweets is really hard, no matter how sophisticated your linguistic and sentiment algorithms are. We're just not there yet.
There's a lot of talk these days about measuring "online sentiment". What exactly is sentiment and how can companies use it? Here is an excerpt from an article that talks about how one such company, NetBase, measures sentiment:
"To assign a sentiment score, NetBase calculates the difference between each company's positive and negative remarks, divided by the overall number of those comments. That generated a "net sentiment" score. Companies that had fewer than 10,000 mentions were excluded."Software applications that help measure sentiment are expensive. If you're a large consumer brand it's a must. It can lead to more proactive customer support, better brand management, and even uncovering sales opportunities. Think of it as a real-time clipping service on steroids.
But most B2B companies and I'm guessing 95% of HR software and services providers don't come anywhere near 10,000 mentions in a lifetime. And most of their online mentions have nothing to do with "sentiment". If you want to know who is talking about your brand most companies will be fine using free services like Tweetdeck to monitor how often their brand is being discussed on Twitter. Similar free services exist for other social network monitoring.
Can businesses make predictions using data from other social networking sites?
In the above referenced USA Today article a sub-headline said "Romney's 1.5 Million Facebook Fans Give Him the Edge". But your total number of "fans" is a result of how effective your marketing has been. Romney's 1.5 million Facebook fans is the RESULT of this success, not the reason for it. So the business value in trying to make predictions based on "fans" or "likes" is questionable. Again, we're just not there yet.
What is valuable - and critical for companies to be doing - is knowing what's trending in their marketplace and discovering the people who are engaged with and talking about the topics important to their business. If you sell workforce planning technologies there is tremendous value in knowing who is talking about topics related to workforce planning on their various social channels.
And this can't be done effectively by just analyzing tweets or the "sentiment" of those tweets. Most tweets don't include a company or brand name or even mention the topic (e.g., workforce planning) by name - but they may link to articles that do mention your brand and discuss topics of interest to you.
These are the people that you want to discover - the ones engaged with the topics important to your business. From there, you can begin to build meaningful relationships that in turn will increase your visibility, your "likes", your "fans", etc. And this is how you ultimately win your elections - and grow your business.
This is what we built SocialEars to do. Allowing you to produce, find, comment on and share your information with the right audiences. The audience in your solar system, not the noise of the outside universe. And this is a representative sample.
To learn more about SocialEars, join us for a webinar on March 9th or March 20 at 1pm - 1:30pm PST.
Labels: Social Analytics, social listening software, SocialEars