- JPMorgan Chase has created an internal tool to make sure its ads don’t end up next to unsavory content on YouTube.
- The company’s proprietary algorithm plugs into YouTube’s API to select “safe” channels for it to advertise on.
- From more than 5 million channels the brand has winnowed the list down to 3,000 YouTube channels that its ads appear on.
- “The model that Google has built to monetize YouTube may work for it, but it doesn’t work for us,” said Aaron Smolick, executive director of paid-media analytics and optimization at JPMorgan Chase.
When some brands’ ads ended up next to troubling videos on YouTube last March, JPMorgan Chase responded by pulling its ads from the platform.
Dissatisfied with YouTube’s slow response, Chase decided to take matters into its own hands and create an internal tool to make sure that its ads don’t end up next to unsavory content on YouTube, only going back to YouTube after testing the tool.
The company developed its own proprietary algorithm in-house, and it plugs into YouTube’s application programming interface (API) to select “safe” channels for its ads to appear on at scale. The algorithm was built by Chase’s internal programmatic and media-buying teams.
“When news broke about ads finding their way next to horrific pieces of content, we paused our efforts and pulled our ads from YouTube,” Jake Davidow, the executive director of media and channel strategy at JPMorgan Chase, told Business Insider. “We wanted to figure out a scalable solution and make sure we got it right.”
The technology consists of 17 layers or filters, which allow Chase to separate what it deems as good or safe YouTube channels from the bad or unsafe ones. One of the filters, for example, looks at the total video count on a channel, which automatically sifts out channels with one-off viral videos. Chase also looks at channels' subscriber counts, the general topics channels focus on, language, and even the comments on different channels' videos.
"The model that Google has built to monetize YouTube may work for it, but it doesn't work for us," said Aaron Smolick, executive director of paid-media analytics and optimization at JPMorgan Chase. "The attention of protecting a brand has to fall on the actual people within the brand itself."
Chase's move reflects an increasing skepticism among major marketers regarding digital advertising, with issues of ad fraud, transparency and brand safety coming to a head in 2017. With their ads ending up next to dicey videos or being viewed by bots instead of humans, many advertisers have begun taking charge themselves by bringing advertising in-house and slashing the number of sites and channels they advertise on.
YouTube in particular has been a conundrum for advertisers. The platform came under fire in 2017 for a spate of incidents where ads ended up next to questionable videos, and brands were not exactly happy with its tackling of the problem. But YouTube is too big for most global advertisers to turn away from, even if they don't have full confidence in it. Users watch an average of 40 minutes a day on YouTube globally.
Chase, along with the likes of P&G and Unilever, has tackled these challenges head on. It started white-listing - or pre-approving sites that its ads run on - in March, culling the number from 400,000 down to 5,000, and about 10,000 at present.
Chase began working on the YouTube algorithm in August and rolled it out in October. From over 5 million channels, it says that it has winnowed the list down to 3,000 channels on YouTube that its ads appear on. Chase says the algorithm has a success rate of 99.9%. The brand also continues to conduct regular manual checks on its channels as well as develop the tool further to make sure that it is foolproof.
"The biggest lesson for us was that we realized that it wasn't a black-and-white conversation with good guys or bad guys, but a gradient," Smolick said. "It isn't necessarily about brand safety, but rather brand appropriateness. That's the next evolution of the debate, with each brand deciding what's appropriate for them and what's not."