“Under the Hood” is a new series of articles focusing more on the technical side of video advertising. In this first of the series, Yuval Fisher, CTO of RGB Networks, a provider of multichannel video solutions, discusses some of the technical challenges associated with combining linear advertising and real-time ad decisioning.
As the media landscape changes, video is being consumed across more devices than ever before. Faced with this growing competition for subscriber attention as well as a limited advertising revenue pie, video operators increasingly realise that they must pursue new, more advanced technologies to broaden the appeal of traditional TV commercials. However, though advertisers and operators alike want to provide more relevant and targeted ads to create more engaging and lucrative experiences, existing technology is limited in its ability to serve ads more selectively or strategically.
We know that a sizeable chunk of TV advertising revenues are based primarily on delivering linear ads to mass audiences – with all viewers watching the same channel, seeing the same ad at the same time. Think of huge broadcast events – like screening the World Cup football final, for example – which attract millions of watchers.
However, a fundamental problem in the existing ad insertion ecosystem is how to address the load involved with making essentially simultaneous ad decisions for huge audiences. Note that this is not a problem with on-demand video, where ad decision requests are naturally distributed; users start watching video – even the same video – at slightly different times, thus distributing the ad decisions over time.
To add insult to injury, the systems that must make all these ad decisions will sit dormant and unused for most of the time, because ad opportunities occur in only a small proportion of the overall viewing time of a programme. So operators must provision significant resources to support huge peaks of ad decision requests and then watch those resources sit unused most of the time. That’s an expensive problem.
This issue has gone largely unaddressed, partially because most addressable advertising today targets on-demand content and partially because linear ad delivery hasn’t really caught up to use the full power of RTB and full user-level decisioning. But is the problem intractable? The specifications used to notify ecosystem components of ad opportunities – e.g. SCTE-35 or ESAM – didn’t anticipate this problem, so a very clean solution is not possible without significant modifications to existing network components, and that’s not realistic. But there are latencies and other signalling opportunities that can be used to reduce the problem and smooth the ad decision requests over time.
The key is to intercept the ad opportunity as early as possible and start the process of making ad decisions immediately, so that these requests can be distributed over as much time as possible. The earliest opportunity to note an ad avail is the initial pre-roll cue that signals that an ad opportunity will occur in a few seconds, and that’s typically detected at the transcoder creating the video feed sent to clients. The transcoder usually would make a request to a component called a POIS, and peeling off the signal at the POIS and kicking off the ad decision processes while the video works its way through transcoding, packaging and delivery can spread the ad decisions over a longer time, changing an expensive and possibly not-viable architecture into one that is manageable – ad decisions that had to be made within a second or two can be distributed over tens of seconds, reducing the load and expense of the ad decisioning ecosystem significantly. The details are a bit tricky, but tricky and tractable is better than basic and broken.