Following the recent revelation on the BBC’s Newsnight programme that ads for the UK government National Citizen Service (NCS), as well as businesses, charities and the BBC were found running next to extremist videos on YouTube, it seems likely that many advertisers will be asking serious questions about brand safety on the world’s largest video-sharing website. One company working on a solution is Veenome, who specialise in analysing various aspects of online video, from the content itself right through to where the player sits on the page. Prior to the Newsnight story, the company recently launched a a new product aimed at analysing and indexing the content found on YouTube channels. Here Kevin Lenane, CEO of Veenome, explained to VAN how the product works and why YouTube’s own targeting isn’t always enough for everyone.
Could you explain how your new YouTube channel product works? Who is the product aimed at?
YouTube Channel Data uses the Veenome indexing APIs (categorization, brand safety and viewability/clutter) to proactively monitor YouTube channels for various data attributes like the subject matter (sports, fashion etc) and brand safety (drugs, explicit content, cruise ship disasters etc) of the actual video content and also records other factors like auto-play, player size and muting. By indexing videos as they are uploaded the channel data is constantly up to date. Data can be accessed in aggregate by channel or by individual video page. Channels are scored based on these factors and raw data is also provided such that brands, agencies and advertisers can choose a channel or a group of channels based on how well their message is targeted and how safe their brand message is.
Is this more about finding content that isn’t brand safe, or is the focus more on providing relevant contextual data?
Its about indexing the long ignored “viewing moment” for the actual videos for both targeting and brand safety purposes. YouTube is a huge source of video inventory but traditional video advertisers can sometimes shy away from it and other popular user generated content because of the risk in the sheer variety of content that YouTube is known for. By understanding what the real video content is about, advertisers can see much greater results by targeting viewers at the most important level – that is “the moment” at which they are watching the video.
For example, if in a given moment my motive is to watch a video about technology, then my interest at that time overrides anything about a demographic to which I am assigned and typically targeted against. This is the product’s key value, which is to target based on that moment while also ensuring that the advertisement is also not associated with the wrong moment. An example of ads being shown at the “wrong moment” would be the BBC beheading video incident from last week. This sort of thing happens all the time.
YouTube has its own methods for classifying and vetting videos. Do their methods fall short?
The YouTube method combines some YouTube indexing as well as information from the actual video provider or publisher. YouTube has some automated controls to prevent broad categories like nudity or copyright infringement and then the video producers can upload subject matter information and metadata.
However, the needs of a video advertiser can be more nuanced and require consistency that they can’t get from the out of the box solution. So from a brand safety perspective, one advertiser might tolerate or even embrace alcohol usage while another may totally object to it. When we look at targeting, the data uploaded by one of millions of various video producers or publishers at a channel is just not consistent or granular enough to really ensure that the subject matter is tracked.
Furthermore, these producers (and YouTube themselves) have their own motivations that cannot be ignored whether this bias in intentional or not. We provide a third party measurement that uses a consistent methodology across all channels and videos which can be relied upon because the information is gathered from image and audio building blocks of the actual videos. Because of our consistent and unbiased approach this method can be applied across many channels and videos, which allows for larger swaths of available vetted inventory for larger campaigns.
Once you have a data set of, for example, car videos. How do you go about integrating that data into the buying decisions?
We do this proactively through partners as well as passively. On the proactive side one of our partners, EQ Works, will actually sell and execute buys based on data that we could provide. For instance, from our constantly refreshing indexing pool of videos we provide EQ with YouTube videos and channels that contain exclusively Do It Yourself (DIY) content and zero non-brand safe videos (so no suggestive, offensive, violent, or drug related content etc.). for a specific advertiser. These specific videos and channels become the actual inventory for the campaign and offer highly targeted, totally brand safe viewing moments at a relatively inexpensive CPM. Our recent campaigns with EQ have been tremendously successful based on the dollars spent for the various success metrics like completion rate, click through rate and engagement rate.
On the more passive side, we provide this information to exchanges and networks who programmatically measure which kind of content matching (ie what advertisers do best with what kind of video content) Once they know this, they either obtain more of that kind of specific video inventory or merely re-route advertisers to the appropriate types of content already within their network or exchange thus improving the success of the campaigns without increasing the generic cost of inventory. This has been tremendously successful with several of our network clients who understand how much the actual video content and that “viewing moment” drive the engagement with video ads.
What type of uplift are advertisers likely to see in terms of how ads perform?
Our first campaigns with YouTube inventory have been more successful then we even thought possible with standard performance metrics like Click Through Rate and Completion Rate more than doubling for the exact same inventory cost. We’ll have some incredible campaign data to release on YouTube and non-YouTube inventory over the next week or two, which describes the truly tremendous dollar value results of this kind of real video content targeting.
