Touching a Nervve: Searching for Advertising Dollars with GPUs
OREANDA-NEWS. August 18, 2016. TV deals are where the big money is at in professional sports. This makes it crucial for companies advertising during broadcasts, and for sports franchises, to understand how much screen time brand logos get when they’re negotiating ad rates.
Nervve Technologies, a Buffalo, New York-based startup, is changing the game of ad negotiation. Its GPU-powered technology allows teams to search hundreds of hours of video in a few minutes, instead of weeks or even months.
From the CIA to the NBA
Nervve is a company more familiar to the FBI and CIA than the NBA and MLB. It established itself by selling its lightning-quick search capability to a variety of federal law enforcement agencies — “all three-letter agencies,” says CEO Thomas Slowe, co-founder of Nervve, which counts In-Q-Tel, the investment arm of the CIA, among its investors.
These agencies are keenly interested in speeding up their efforts to comb through security footage, social media posts and still images. They need a fast and precise technology.
Nervve’s technology can scan a data-rich image, say a satellite photo of Manhattan, quickly search for and find its target — perhaps a newspaper — and gather all relevant exposures for an analyst to quickly review, Slowe says.
Government agencies aside, Nervve may have found its footing in the lucrative market of major professional sports organizations. Up to now, these organizations have relied on tedious manual processes to determine how much exposure their advertisers get on TV.
Nervve can quickly calculate how many times advertisers’ logos appear during a broadcast, the duration of the appearance and its size, whether it’s on stadium signage, jerseys or even the sneaker of an athlete. Nervve is drawing sports executives because “we automated a process that, to this point, has not been automated,” Slowe says.
Providing Negotiations Leverage
Data extracted by Nervve is proving crucial to negotiations on rates. The New York Mets and other Major League Baseball teams have been using Nervve to analyze their game footage. More recently, the National Basketball Association turned to Nervve to help determine how much revenue adding patches with advertiser logos to player uniforms would create.
Subsequently, several NBA teams are now using Nervve to determine how often logos, whether on patches or arena signage, are appearing during broadcasts. Advertisers are using Nervve, too, Slowe says.
Teams and advertisers don’t actually buy the technology. Rather, they buy marketing reports and raw data summaries as a service. Nervve plans to adopt a subscription model going forward, Slowe says.
GPUs Part of Nervve’s DNA
GPUs have factored into Nervve’s search platform since the company’s founding in 2011. Co-founder and CTO Jake Goellner is an expert in CUDA, NVIDIA’s parallel computing platform and programming model. To accelerate the performance of Nervve’s technology, which is based on massively parallel processing, Goellner and Slowe ported it to CUDA.
Eventually, the company added GPUs to speed up its search algorithms. It currently uses a combination of NVIDIA Tesla K80 and K20 GPU accelerators to turn pixels into mathematical representations for high-speed comparison. Slowe estimates that the process unfolds five to 15 times faster with GPUs.
“It definitely gives us a valuable lift on what we push through our systems,” Slowe says.
Nervve also has begun publishing what Slowe calls “syndicated reports.” By releasing summary sets of data on sporting events of great interest — most recently Wimbledon and the MLB All-Star Game — Nervve hopes to expand its market.
“Anyone who wants more data that’s related to them can follow up,” says Slowe.