For most businesses, it would be a PR nightmare: A major technology company had just released a software update that caused its tablet to crash. Consumers were beginning to voice their complaints online. But their nascent grumblings were spotted early by Blab, a growing analytics startup that hunts for timely topics before they trend on the web. Blab warned the tech company that the conversation would escalate from murmurs to shouts, primarily on Facebook and news blogs, in approximately 40 hours. "So we gave the customer a heads-up: ‘You have 40 hours to put ads in place before all of the blogosphere starts talking about this issue,’" recalls Blab founder and CEO Randy Browning. "So the marketers at this company knew they had 40 hours to mitigate this issue that’s coming."
Predicting the future is just business as usual at Blab, which sniffs out budding conversations across the social media world up to three days before they hit the mainstream. The Seattle-based company uses its crystal-ball findings to help brands anticipate what will resonate with their audience 72 hours in advance, enabling them to roll out efficient marketing—in the right place, in real time. "We bring the future into perspective," Browning says. "We give brands tomorrow’s newspaper today."
Browning, a veteran of the advertising and marketing worlds, was increasingly drawn to the marketing potential of social media. He saw social networks as virtual block parties where users traded recommendations and opinions on brands relevant to their lives. The question for brands has been how to participate in the party. When companies spend precious ad dollars, they want to be in control of how their message is delivered, Browning says. But online communication moves so fast that these messages are often outdated by the time they reach their intended audience. "You have brands losing the power to influence people in their decisioning," he says. "That became a disruptor for our industry. No longer can we rely on one-way communication—it has to be two-way. And that really has to be in real time."
Real-time marketing made a splash during the infamous half-hour blackout at the 2013 Super Bowl, when Oreo tweeted an ad reminding viewers, "You can still dunk in the dark." To capitalize on the moment, the company had a full social media team poised to react to any viral event within minutes. But not every brand has the flexibility for such a speedy response. That’s where the value of predicting the future comes in.
Browning put together his engineering team and founded Blab in 2012. But the Blab crew quickly hit a wall: In trawling reams of data online, they found conventional Natural Language Processing too limiting; using search terms and hashtags was slow and costly, and it ignored relevant conversations that didn’t include keywords. So Browning and CTO Stefan Papp decided to rely on statistics rather than linguistics. They created a proprietary system, BlabPredicts, that classifies conversations based on common patterns. The platform uses these patterns to predict when, where, and at what velocity chatter will unfold—and how loud it will get. "Instead of looking for what I think is important, I’m discovering what people are talking about," Browning says. "We can accurately, to the half-hour, tell you when a conversation will take off."
For example, Blab recently worked with a detergent company to research the topic of scent. "No one speaks with ‘#scent’—that’s just not English," Browning says. "So the search approach is not the best approach. But we start with scent, and our engine learns all of the associated words, terms, and phrases, and we start to find the conversations relevant to those terms. We can find conversations that don’t even have the terms ‘scent’ or ‘smell’ in them." Over Halloween last year, Blab helped the detergent company zero in on a fledgling problem: kids having trouble washing the musty odor out of thrift-store costumes. "It was great because this customer would have never thought of that," Browning says.
Amid a sea of social-analytics tools, Blab differentiates itself in several ways. Instead of relying on historical data, the company gleans insights from its complex conversation-modeling system. BlabPredicts makes about a million predictions every minute, while sifting through more than 100 million conversations per day. The platform can track a single conversation across a dizzying array of formats, including tweets in any language ("even Klingon," Browning says), a smiley face on Facebook, a photo on Instagram and a three-second video on YouTube, for instance. Blab then predicts the discussion’s evolution so clients can market the right messages to a receptive audience, or manage a bubbling crisis before it erupts. Overall, the system has an accuracy rate of 70%.
Blab might not be what most companies use for its basic social media analysis, such as measuring ongoing brand sentiment, says Julie Hopkins, a research director with Gartner, an IT research and advisory firm. "Where Blab fits in is if companies have the ability to move quickly and create content and messaging based on individual, segmented feelings within your audience, this is an extra leg-up," says Hopkins, who included Blab in the recent Gartner report, "Cool Vendors in Social Marketing 2014." "Marketers are doing everything possible to get their brand in front of the right customer, at the right time, with the best possible offer. The ability to turn on a dime can offer an additional bump."
So far, Blab has raised $5.85 million in funding and has a client base of 10 "global brands" that Browning is keeping confidential. The company launched version two of BlabPredicts last month, and Browning is now focused on building partnerships with media and ad agencies. "For the first time," he says, "we have a predictive data set for media agencies to be able to place media." The excitement in Browning’s voice is palpable. Predicting the future of the Internet might be a day’s work for him, but that work is still a thrill: "It’s super awesome," he says.
[Crystal Ball: Anneka via Shutterstock]