Photographer: Getty Images

The Company That Cracked Image Search

Nervve’s image-recognition software is used by everyone from the NBA to U.S. intelligence agencies.

Advertising is coming to NBA team jerseys next year. In April the league’s owners voted to allow 2.5-inch by 2.5-inch sponsorship patches on the front left shoulder of game uniforms. It’s novel real estate: No other major U.S. team sport permits such displays, though soccer clubs here and around the world have sold sponsorship on shirts for decades. To figure out how much exposure the new patches will get and how valuable they are, some NBA teams are turning to Nervve Technologies, a Buffalo-based company whose visual search software is used by U.S. intelligence agencies.

Nervve moved into sports last July, when it linked up with Wasserman, a talent agency and marketing consultant that specializes in sports. Earlier this year, according to Amy Brooks, the NBA’s head of team marketing and business operations, the league asked Wasserman/Nervve and about 10 other consultants to demonstrate how they could help teams evaluate sponsor patches. At the All-Star Game in February, both teams wore Kia logos on their shoulders. It was, in part, a test run. After team owners approved leaguewide patch sales, the NBA recommended two of the consultants, Wasserman/Nervve and Repucom, a global sports-research company. “Nervve really stood out from the speed perspective,” says Brooks. A dozen teams, including the Cleveland Cavaliers and Orlando Magic, are working with Wasserman/Nervve to analyze the market for the patches.

Nervve’s software can track how often, where, how large, and for how long a sponsor’s logo appears to people watching the game on TV or any other screen. The software can scan an hour’s worth of footage in five seconds or less. “On a single server, we can scan an entire season of Yankees games in under two minutes for a particular brand,” says co-founder and Chief Executive Officer Thomas Slowe.

Before Nervve, says Zack Sugarman, Wasserman’s head of properties, his analysts sat for hours watching sporting events and logging when and where sponsor logos appeared on stadium signs, blimps, hats, and so on. “We’ve been doing measurement and evaluation for years,” says Sugarman. “Nervve just lets us do it more accurately, faster, and at larger scale.” Wasserman uses Nervve’s raw data to help leagues, teams, and advertisers determine the visibility and worth of in-game ads—and where to place them.

Slowe has been studying machine learning since his undergraduate days at Rutgers University. After receiving a master’s from MIT’s Media Lab and serving as a research scientist at Ericsson, he worked at a motion-detection startup in Northern Virginia. In 2003 he founded a company that analyzed facial expressions.

With co-founder Jacob Goellner, Slowe started working on the technology in 2011 for what became Nervve, aiming to serve U.S. intelligence agencies. From his previous ventures, he knew the intelligence community was buried in surveillance footage. “An analyst will get a terabyte drive of a hundred or a thousand hours of video on their desk each morning,” he says. The automated systems available to scan for threats typically required rooms full of servers and operators with advanced degrees. Nervve created a drag-and-drop interface that allows a minimally trained user to find objects quickly and accurately, without a lot of hardware.

The company spent two years figuring out how to isolate patterns in pixels without relying on the long strings of yes-or-no questions usually found in machine-vision algorithms. Slowe uses the metaphor of a jigsaw puzzle: Instead of picking up pieces one at a time, Nervve scans a jumble all at once and makes guesses, leaving yes-or-no decisions to the very end.

Late in 2012, Nervve began testing the system for what Slowe calls “a lot of three-letter agencies,” without specifying them. In-Q-Tel (IQT), the investment arm of U.S. intelligence agencies, became one of Nervve’s backers in 2014; it hasn’t said how much it’s invested. Nervve licenses its software to government contractors, who run the searches themselves. Slowe says the government uses the technology, for instance, to look at video collected by drones.

Once the government business was up and running, Slowe turned his attention to the private sector and began soliciting media executives. Wasserman saw the potential value right away. The agency, says Sugarman, has already used Nervve’s technology to analyze brand exposure in football, baseball, Nascar, and Kentucky Derby telecasts.

The first patch sale came on May 16, when the Philadelphia 76ers announced that ticket seller StubHub would add its logo to the team’s jerseys. (Wasserman doesn’t work with the 76ers.) StubHub will pay about $5 million per year for the space, according to an NBA official with knowledge of the deal. Sugarman expects some teams to charge three times as much.

The bottom line: Sports teams and leagues are using Nervve’s image-recognition software to assess the value of in-game advertising.

Before it's here, it's on the Bloomberg Terminal. LEARN MORE