Twitter Feeds on Netflix Prompt Stock Service
The torrent of posts on Twitter Inc. last year about Netflix Inc. (NFLX) might have tipped off investors to stock swings that erased about $8 billion from the movie-rental company’s market value in three months.
Topsy Labs Inc., a site that searches data on the microblogging service, analyzed tweets during a period around Netflix’s decision last September to split its DVD and streaming businesses. By examining posts such as “just canceled my Netflix subscription,” Topsy was able to pick up signals of a coming drop. Now, it’s taking those results to help build a new service to aid investors in predicting other stock moves.
“Information dissemination is now largely not through publication, but through conversation,” said Rishab Ghosh, co- founder and vice president of research at Topsy, based in San Francisco. “We’ve actually been able to show that you can take that data, put it through processing, and relate it to the market return for a specific stock.”
Topsy is one of several companies, including WiseWindow and Derwent Capital Markets, that use so-called sentiment analysis, a cutting-edge field where companies examine the chatter on Twitter, Facebook Inc. and blogs to help them predict stock movements, market trends and the success of new products.
Social media’s growing influence throughout the world, such as the protests in the Middle East during the “Arab Spring,” has made sentiment analysis an important tool in other areas, such as the measurement of political risk. It also may have broad applications in gauging voter attitudes during, for instance, an election year.
Sentiment about Netflix was found by tracking negative and positive comments on Twitter in a given day. For example, “an apology from your CEO won’t win my business back” would be considered negative, while “Netflix split a good long-range move for the company” was positive.
The company also used baselines to adjust for distortions in data, such as the time of day. For example, the Topsy analysis discounted negative comments on Monday mornings, when people tend to be less upbeat.
“It is data and news generated from the ground level,” said Erin Collard, an adviser to Topsy Labs and a portfolio manager at Armored Wolf, a fund with about $750 million under management. “It’s real time, and it’s way more predictive, and way ahead of the curve before it even reaches mainstream media.”
The company’s analysis of Netflix showed that of all the factors that could influence a stock’s direction, sentiment on Twitter on a given day captured more than half of the stock’s move the next trading session.
Access to Investors
Topsy Labs, which has about $30 million in funding, plans to offer access to Twitter sentiment analysis later this year to investors looking for guidance on stock-price moves, Ghosh said. In an agreement with Twitter, the company has full access to the messages of the service’s more than 100 million users, he said. That means it sifts through tens of millions of tweets a day, not just small samples, and can look at historical data it has had for years.
“In order to best understand and put into context what’s happening today, it’s helpful to understand what’s happened as far back as possible,” said Lou Kerner, an analyst at Liquidnet Holdings Inc. in New York. “That’s an advantage.”
Still, the Twitter data can’t be a complete reflection of what’s going on in the world because there are other outlets, such as social-networking giant Facebook, where people can express opinions, said Jake Wengroff, a social-media analyst at Frost & Sullivan in San Antonio, Texas.
“Only looking at Twitter is going to have its limitations,” he said. “If people have something to say, Twitter might not be their first choice.”
Topsy previously looked at how sentiment about Apple Inc.’s iPhone 4S indicated that sales might be better than what analysts initially expected. On Oct. 4, the day that iPhone was released, initial comments about the new device were more positive than negative, according to Topsy.
To contact the editor responsible for this story: Tom Giles at firstname.lastname@example.org