Topsy Labs Inc. released a report claiming its model was able to predict a drop in Netflix's share price after it decided to split its DVD rental and streaming video services by tracking phrases like "just canceled my Netflix subscription."This week
It's arguable whether investors really needed a sophisticated sentiment measuring analyses to predict Netflix's shares would drop after what has been called the worst business decision since the introduction of "new" Coke in 1985. But social media sentiment analysis is growing more sophisticated and may soon become a key component investors look at before making a decision to buy or sell stock.
"In future it may very well be unwise to not take some kind of sentiment analysis into consideration," said Neil McGovern of Sybase, a company that provides its complex events processing technology to banks and securities firms, "But at moment it's not make or break variable."
How Twitter Beat The Mainstream Business Media On iPhone 4S
"People don't really tweet about stock ticker symbols, at least not in any predictable or useful way," said Rishab Aiyer Ghosh of Topsy. "But they do talk about companies and products."
Ghosh pointed to the launch of iPhone 4S, which was initially greeted with lots of negative reviews from the mainstream media. Those negative articles fueled a steep decline in Apple's share price on the first day iPhone 4S was available. But on social media, early reviews were mostly positive with normal people tweeting about how they liked and wanted the latest version of the iPhone.
"Ten days after the launch Apple announced they had the biggest backlog for a product in its history," Ghosh said. "The day of the release the Twitter chatter was more positive than negative: as an analyst, if you had known that, you would not have sold your Apple shares and you would have made a lot of money."
Later this year, Topsy will begin marketing its historical and real-time data to trading firms and equities analysts. Ghosh, like everyone else interviewed for this article, stressed that the social media sentiment data was just one component used by Wall Street to build trading models.
"The big-picture social media content data we provide can augment, if not totally replace, the news media content," he said. "One of the advantages of the news media is its curated and it doesn't have a lot of noise; what we do is take the noise out of the social media sentiment data."
Do's And Don'ts Of Trading On Social Media Sentiment
How much of an advantage does social media sentiment give a trader? Last year WiseWindow commissioned an independent researcher to track its stock-picking model as it followed four stocks - Ford, General Motors, Southwest Airlines and American Airlines - for the first six months of 2011. The model imporved performance between 30% and 48% on an annualized basis.
"One area we haven't really been able to get into these models is consumer sentiment: survey data," said Marshall Toplansky of WiseWindow. "If you do a traditional survey, you may be able to get 1,200 or 1,500 comments. With social media we can get tens of millions of comments per month."
For Toplansky, there are several keys to using social media sentiment to predict movement in an individual stock, including:
- Using data from several different platforms, including Facebook, blogs and even comments left on YouTube videos. "Twitter on its own is not a leading indicator," he said. "You have to take into account all of the different forms of social media."
- Tracking conversations on a stock or a $ thread on Twitter won't cut it either. Toplansky found that the biggest predictor of a stock's price are comments about product quality, which often don't mention a ticker symbol or even the company by name.
- That can mean tracking up to 2 million social media messages in a month, as opposed to the few thousand WiseWindow would uncover if it only focused on discussions about a company's share price. "You continually have to go out and look for patterns -- it's almost a 24-7 kind of thing," Toplansky said. "You have to do the math over time and see what trends evolve and emerge."
Strategy Doesn't Work For Individual Investors - Yet
And therein lies the problem for individual investors, as well as many investing firms: Sybase's McGovern said the upfront costs and the lack of a proven track record make it difficult for most investors to justify adding a sophisticated sentiment analysis component to their trading models.
"The reason all of this is appealing to people is there's a general understanding that shares obviously don't just move on the underlying aspects of the company," McGovern said. "The fact that stocks can go up 5% or10% day, which is unusual but not unheard of, can often be because of rumors in the market. This all seems to make sense to people in the markets as a way to be able to tap into those rumors and help their short-term trading strategies."