Data Firms Quietly Profit in Wake of Twitter IPO

Companies that sell and analyze data from tweets have been thriving since the social media outfit went public.


It’s not just Twitter shareholders who have been smiling since the social media outfit’s November 7 listing on the New York Stock Exchange. As the stock has surged past $50 from its opening price of $45.10, several less prominent businesses that sell and analyze the data from the so-called fire hose of tweets have been reaping the benefits of their association with the San Francisco–based company.

Apple acquired Topsy Labs, one of four firms licensed to package and resell the full fire hose of historical and real-time tweets, for a reported $200 million in early December. Shortly afterward DataSift, another licensed reseller, announced that it had raised a further $42 million in funding, to bring its total to $72 million.

NYSE itself has gotten in on the act, partnering with Chicago-based data cruncher Social Market Analytics to push out alerts to its customers when the sentiment around particular stocks changes. SMA’s algorithm extracts scores from social media noise that give subscribers early warning of imminent shifts in price, the company says. “It’s not what’s being said but who’s saying it that counts the most in terms of stock performance,” explains president and CEO Joe Gits, adding that of the 230 million monthly active Twitter accounts, the algorithm focuses exclusively on the 400,000 most influential market makers.

The family of “indicative” tweets — those that can be used to assess stock sentiment — is growing by 10 percent every month, according to Gits. Even as the social media platform grows, the conversation on Twitter is becoming more, rather than less, useful for the markets.

SMA’s main score grades stock sentiment on a scale from –5 to 5, broadly reflecting the standard deviations from the mean in a normal distribution, with the score being revalued each minute based on the evolution of noise from indicative tweets. The firm’s main claim: Changes in that score often lead changes in price. For example, the sentiment score for Apple increased from 1.5 (weakly positive) to 2.5 (highly positive) more than three hours before investor Carl Icahn revealed, via a tweet, his “large position” in the iPhone maker on August 13; trading off the sentiment signal alone would have allowed SMA users to preempt the price spike that followed Icahn’s intervention.

“There’s alpha in the sentiment signal,” contends Tom Watson, vice president of global market data at NYSE Technologies, the data and technology division of NYSE Euronext. Watson points to the superior returns generated by SMA’s backtests, which show that by simply trading on all S-scores greater than 2 for companies in the Standard & Poor’s 500 index between December 2011 and August 2013, users would have generated a cumulative return of 39.05 percent with a Sharpe ratio of 1.72; over the same period the index itself returned 13.75 percent with a Sharpe ratio of 3.77. Although Watson concedes that this alpha will eventually be arbitraged away as SMA sentiment scores become more widely used, he thinks they will remain useful indicators. “Twenty years ago no one was plugging earnings estimates into their equities models, but now everyone does,” he says.

Other companies have taken a less sentiment-focused approach to mining Twitter for market-moving information. Whereas SMA uses a slice of vetted “influencers” to generate a series of general sentiment scores, New York–based Dataminr traverses the entire fire hose of nearly 500 million daily tweets with a sophisticated multipart algorithm to zero in on the single tweet with the potential to move the market.

For Dataminr the obscurity of the tweeter is no barrier to a particular tweet being relevant to the markets, so long as the credibility of the information can be verified, a process its proprietary algorithm controls for. The firm, which raised an additional $30 million earlier this year to bring its total funding to $46.5 million, doesn’t give its users a score the way SMA does; it simply delivers the raw tweet.

In early November, for example, three minutes before the news hit financial wires and two minutes ahead of the subsequent sell-off, Dataminr alerted clients to a tweet stating that troubled Canadian telecommunications company BlackBerry would replace its CEO. BlackBerry’s stock price eventually declined 20 percent in response to the announcement.

NYSE road-tested Dataminr and SMA before inking its partnership with the latter, the exchange’s Watson notes. “SMA was simply more closely aligned with our model, where we provide our customers with information in a feed,” he says, adding that both products can be useful. “There’s a lot of different ways you can use social media.”

But Twitter sits at the center of the movement to incorporate the unstructured noise of social media into trading strategies. “Two years ago there was a much higher dispersion in the sources of online data” looked at to generate meaningful trading signals, says SMA’s Gits. “Now people just want Twitter.”

On the morning of Twitter’s IPO, CEO Dick Costolo, who’s faced repeated questions on the company’s revenue model, told CNBC that he has a “whole set of strategies” around “bridging the gap between the massive awareness of Twitter and deep engagement on the platform.” Companies like SMA, Dataminr, Topsy Labs and DataSift have shown that, for now, there’s value to be had from living in the shadow of the fire hose.