Digital video’s multi-year climb as the fastest growing medium in both consumption and advertising continues unabated. The average viewer now watches more than an hour and a half of online video content per day, and Google projects that 85% of total digital consumption will be video by 2020. More video content is uploaded every 30 days than the major U.S. television networks have created in 30 years.
The advertising trajectory for digital video will leap another 17% in 2018 to $15.4 billion, according to eMarketer, with 87% of marketers promoting with video content.
The explosion in user-generated video content has
created challenges for managing video advertising buys. According to Advertiser Perceptions, the main concerns of advertisers looking to achieve scaled video programs include:
-- Running in brand-safe environments
-- Only paying for completed ads
-- Being guaranteed viewable ads, and
-- Having control over the content in which the ads run.
Some would say these goals are perfectly attainable as long as you don’t mind paying a premium. For instance, Google Preferred offers an option that vets all placements on YouTube to ensure none will land in unsavory places. However, this service is largely a manual process and comes at a price many don’t care to pay.
Other companies use a combination of manual plus technology to create white-listed video buys beyond YouTube, but these are also premium priced.
Given the vast volume and splintered fragmentation of video inventory, it makes sense that AI technologies will be used to sift through mountains of content to identify the appropriate contextual placements for brands, which matters more now than ever.
Understanding context, after all, is one of the things AI is good at. By tapping into deep learning technologies, advertisers can target the categories of video content that make sense for their brands, and then optimize toward video content that resonates with their customers.
There are several companies applying AI to video inventory to achieve brand safety, efficient targeting and more completed views at scale.
Uru (recently acquired by Adobe) is a highly advanced computer vision technology that can look across millions of videos to identify alarming imagery brands want to stay clear of (guns, nudity, etc.), automatically blocking placement near offensive content. It is positioned mainly as a brand-safety tool, but can also be used to target desirable video content that brands want to associate with.
This kind of video targeting is new and requires a different lens, but advertisers are increasingly using similar strategies of targeting photo images with platforms like Instagram,and GumGum. Uru is just a few steps beyond.
Another platform, Pixability, uses AI to weed out undesirable video content and home in on the right context for brands across YouTube, Facebook, and Instagram. It uses machine learning techniques such as semantic analysis and word vectorizations to analyze the metadata associated with each video, including keywords, descriptions, categories, transcriptions, and comments. AI is applied to curate and package the kind of videos brands like to use for targeting, such as fashion or sports.
“We view brand safety and suitability as subsets of contextual targeting,” says David George, CEO of Pixability. "Advertisers want to appear in suitable contexts that maximize brand impact, while avoiding contexts that don’t align with their brand promise."
AI is also answering the pain point around completed video views. Currently, accepted completion rates of video ad views vary -- some as low as 3 seconds, which is obviously not ideal.
A company I advise, ViralGains, offers 100% completed views of video ads at scale, charging for a cost-per-completed view (CPCV). Videos from brand-safe sites and applications outside of Google and Facebook are mined using AI, which makes it possible to produce more completed video ads.
“The general machine learning model initially used to optimize for engagement becomes displaced over time by the deep learning AI algorithm, which estimates whether a person will engage with a particular video,” according to ViralGains CEO Tod Loofbourrow. "The AI gets smarter and outperforms the generic machine learning model.”
In this way, the algorithm can buy across variable cpms without needing to forego some of the higher-priced inventory, as the result delivered is in completed views at the agreed CPCV – the ultimate KPI many advertisers are seeking to demonstrate campaign success.
Video is too big and too important for advertisers to let challenges get in the way of its value to advertisers. AI solutions such as these are clearing the way for more effective and worry-free programs.