Can social media predict the future?

We all know reporters love to prognosticate, their opinions respecting no temporal boundaries. But now a whole other industry has grown up that lives off their stories, trying to predict future events based on today’s news coverage. Companies like RavenPack have built a business by sifting through news stories in near real-time, looking for the sentiment of coverage related to publicly-traded companies. They claim this information can be profitably used to predict fluctuations in stock prices.

Predicting stock price moves based on current media sentiment has logic to it – readers are also traders, so if we see a string of gloomy stories about IBM’s recent product announcement, then some of us will likely dump the stock. And our likelihood to take action is proportional to our trust in the news source: If we see negative reporting in the Wall Street Journal that outweighs positive opinions on iLoveIBM.com.

Now we have social media and the prediction game has got a good deal more complicated. Researchers at the USC Annenberg School have been working hard to see how social media traffic can be used to try and predict future events. They started by looking at new movies and tried to correlate Twitter messages to first weekend receipts. They managed to get very good at estimating this, often better than industry experts, and what they found has broad implications for social media marketing.

The Annenberg team discovered that the best predictor of box office outcomes wasn’t the volume of traffic related to a movie, but the net sentiment expressed. In other words, quality beats quantity in predicting outcomes. A new movie might get a lot of buzz, but that wasn’t predictive of making a lot of money. This confirms what companies like RavenPack have long known: it’s the tonality, not the totality of coverage that really matter.

The takeaway from this? We need to be very careful how we measure social media outcomes. Measuring retweets, mentions and the raw volume of coverage for your brand just isn’t enough. Yet this is what most people do today.

Annenberg, in co-operation with the LA Times, are now trying to see if they can use their system to predict the outcome of this week’s Oscars. If you believe them, we should expect a big upset: The winner for best movie will be Midnight in Paris. Here I’ll make a prediction of my own: Much as I liked it, Midnight will lose.

Remember, I said that being able to predict stock fluctuations worked because readers are also often traders; the same logic does not apply to Oscar voting. The Oscar outcomes are based on the opinions of a mere 5,700 members of the Academy. Winning others awards such as BAFTA are a much better indicator of Oscar success.

The lesson here is that listening to everyone’s opinions is often a mistake. Instead, we need to target on influential audiences – potential customers, shareholders, employees – and understand clearly how they impact our brand and business.