Monetising video’s long tail inventory is a tricky business. Brand safety is paramount in the video world, but even the most scrupulous of ad networks and exchanges can find ads their ads occasionally end up on the wrong side of the tracks. However, one of the ways ad networks and exchanges can assess the quality of their inventory is to work with a company like Veenome, who analyse the footage and analyse it on a frame by frame basis, whether you’re trying to find horses, dispose of unwanted penises, or track down a company logo.
Veenome was founded by Kevin Lenane, CEO, who explained to VAN how his company’s technology works. “We take visual video and some audio, and translate that into machine-readable data such as categories, tags, objectionable content etc. So we can manage those processes at a large scale and that data can be used for a variety of purposes, including things like campaign audits and analytics, CPM-lift, SEO and trying to make the video content more findable,” he said.
“Our core business is built around servicing ad networks and ad tech, such as AdTheorent, an RTB-enabled mobile ad network, as well as some advanced monetizing publishers such as Videofy.me and Ugroove.We also have some clients who we’re unable to talk about because the work is of a legal nature.”
So what’s fuelling the need for services like Veenome’s? Lenane explains, “I think the main problem is when an ad network doesn’t know as much about their inventory as they’d really like to. That lack of knowledge carries risks that can make brand advertisers very uncomfortable, so it helps that we’re able to shed some light on what the inventory contains and help them police it. That way the network can deliver the best quality advertising and not get black-listed because there’s some content there that they’re no aware of. Screening inventory without technology like ours is a laborious process for networks because it’s so difficult and time consuming to manually check through video content.”
Veenome’s tech can also be used for segmentation, so networks can quickly sort through their inventory and place each video into the correct IAB category. “With AdTheorent we’re going to be using our technology to assist with enhancing their real-time targeting. So they’re going to provide us with the URLs of videos with videos on them, and we can then index them according to IAB categories. So for example, by being able to classify something as a politics video, AdTheorent can then segment their inventory more effectively, which of course raises the value and appeal of that inventory to an advertiser whose ads perform well in that category.”
But how reliable is the analysis? Doesn’t automated recognition of human content, whether it’s audio or images, get it wrong at least some of the time? Lenane is confident that Veenome delivers accurate results, with just a few minor caveats. “It’s tough to answer the question of how accurate the tech is in every scenario, but on the whole it’s really accurate. For example, for things like assigning content categories, it’s over 99% accurate. For things like objectionable content, again it’s over 99% accurate.”
However, things get a little bit more difficult for Veenome when it comes to tagging video content i.e. assigning meta-data to the content to make it more discoverable. “It depends on the type of thing that we’re doing, because for example if you’re asked to tag a video, there’s an infinite number of things in any scene that you could potentially tag, so you have to decide what’s relevant. So we would usually aim to tag either the most obvious thing or something from a specific list. Choosing those things wisely is the hardest thing about tagging,” said Lenane.
“Then we also have to put objects into actions that have a context, so if we see a car we want to be able to say it’s a driving scene, or if we see a horse we want to be able to work out whether it’s a sports scene or a farming scene. But each publisher or network will have specific needs, and it might be as simple as having to pick out a brand logo,” he added.
