Dataminr, a New York startup that has raised $50 million and caters to hedge funds to emergency responders, is growing its footprint into media and probably other verticals because it has a knack for alerting customers to actionable tweets in a sea of noise.
The company has a bevy of examples where its service has been critical to making money and saving lives. Dataminr this year tracked the Boston Marathon for the City of Boston, flagged a shakeup at BlackBerry for traders minutes ahead of the actual news and started telling the tale of a train derailment in Lynchburg, Va. 46 minutes ahead of news reports.
As a strategic partner of Twitter, Dataminr has direct access to the company's firehose and finds relevant information to deliver via alerts on the desktop, email, or in a business system. I couldn't help but wonder if Dataminr's business would make more sense as a Twitter monetization engine than advertising. My other thought was "Gee, Twitter should just buy these guys."
We caught up with CEO Ted Bailey in his New York City office to talk shop. Here's a look at the highlights and key points.
Why build a business on Twitter's data? Twitter has taken a few hits of late as worries about engagement have surfaced. The big concern is that Twitter may not be as sticky as a site like Facebook. Bailey thinks these concerns are unfounded and points to another barometer for Twitter---the velocity of people broadcasting on-the-scene events via the service. "The distinct quality of Twitter is our life blood. There's a dimension of Twitter that's on the front line of open and real-time information. It acts for us as a real-time sensor network," said Bailey, who added that Twitter's secret sauce since inception is that people publish things as they see them in the real world. "When it comes to seeing something in the world and publishing it, Twitter dominates," said Bailey. For Dataminr, Twitter's data set is perfect for spotting trends early. As for engagement, the only thing that may matter for Twitter is that the infrequent user in Kansas broadcasts that factory fire or breaking event. That on-the-scene Tweet is immensely valuable to Dataminr.
"I have confidence in Twitter as a data set. I suspect over time there will be other platforms, but that user behavior just doesn't exist anywhere else right now," said Bailey, who adds that Twitter is also a strategic partner with Dataminr. After all, Dataminr and Twitter did work together to create an early warning system in partnership with CNN.
Industries and use cases. Bailey's hope is that the Dataminr tools for CNN will be used for other media companies. Dataminr for news is expected to be broadly available in the summer. The two mature verticals for Dataminr are public sector entities and finance firms such as hedge funds. Both industries highlight the Dataminr value prop. For instance, if Dataminr can surface a relevant Tweet that gives a trader a three-minute advantage on a stock he's trading, that's real money. For the public sector, Dataminr can comb through Twitter to surface events that first responders will need to know about whether it's a natural disaster or public health issue. "There are 500 million tweets a day and there are only a handful throughout the day that are relevant at a particular moment," said Bailey. "We need to deliver those messages at the right time with the right analytics."
Another likely frontier for Dataminr would be risk management for enterprises. Companies all have to manage reputation risks and a tool like Dataminr could be handy.
Integration. Dataminr's other key value proposition is that it will integrate its services into a company's workflow and systems. For media companies, Dataminr would have to integrate into mostly home-grown content management systems. In finance, Dataminr has to integrate into trading systems. On the government front, Dataminr has to plug into a variety of systems. "What we do is specialize in complex sales and services. It's about consulting and integration and working deeply with the customer," said Bailey. When I asked Bailey whether some systems were easier to integrate than others, he quipped that "it's all a pain to integrate" and "most applications are custom." Bailey said the integration pain, however, is necessary because Dataminr aims to deliver the right Twitter data in a workflow that's relative to an individual person. For instance, a tweet that appeals to a rapid-fire trader wouldn't apply to a fund manager with a five-year time frame to invest.
Given the integration work required upfront, Dataminr's business model revolves around subscriptions and services.
The algorithm (in layman's terms). Dataminr's algorithm is a machine learning process that tries to make predictions. Twitter activity is detected and categorized based on interrelationships, said Bailey. "People are sensors out there and if you look at the Twitter archive over time there's a flow and pattern for every story there has ever been," said Bailey. Dataminr takes that pattern as a story breaks, goes local and then national. A developing stream of tweets will be evaluated by Dataminr and compared to other stories that went viral and discussed by influencers. Dataminr also looks at news feeds, brand mentions, sub brands and scrapes the Web to define concepts and put context around Twitter data. For instance, if there's an explosion in the Midwest Dataminr's algorithm would look at a tweet to see if it was in the news at the moment. A tweet would be more valuable if the news wasn't on the wires yet. Dataminr tries to nail the pre-news cycle for events and the most valuable human sensor may be a person with 200 followers in a region that isn't a Twitter power user. "If we can conquer that group — mainstream people who are participating in events at distinct times — we will have the right signals," said Bailey.