Let’s build our systems flexible, yet realistic based on the data we're seeing, and then we can start to move to the real implementation phase
We’ve been hearing plenty about the various SCTE 130 parts. Then there are the various extensions to SCTE 130, and after that the integration efforts. I have no doubt that all of this will end up helping the industry move forward in a common way, but before you get too frustrated at the pace, let me talk a bit about a few of the pre-130 successes for VOD insertion that are out there. After all, standards follow innovation – so let’s get some of the early innovation results.
We’ve seen a significant focus on time-shifting, from Start Over to RS-DVR. One of the time-shifting applications that I think is particularly interesting is Cox’s MyPrimetime. The key to this service is that the advertisements are replaced from the original broadcast program and inserted on-demand, on the fly, using the SCTE 130 equivalent of an ADM (VOD ad insertion system), a CIS (content management) and an ADS (campaign management).
What happened? On Turner content alone, Cox received an estimated 10 percent uplift in viewing within the first three days of MyPrimetime views. Twenty-nine percent of viewers would not have watched the shows at all had they not been available on MyPrimetime. There are a lot of great statistics that Cox released, but these two show that there is a real business here. Given that ads are replaced in MyPrimetime content, that equates to a 10 percent ad uplift, with 3 percent being new customers altogether.
Using similar components that are described for the Cox implementation, Sunflower Broadband sold out its entire inventory of spots in one day. Dynamic VOD advertising generated a 10 percent increase in ad clients and increased ad revenue significantly.
Finally, let’s look at the Virgin Media trial that ran in London. The key result was that there was no negative impact in viewing patterns for the addition of video-on-demand advertising.
As with print advertising, placement is everything. Prerolls are viewed approximately 95 percent of the time, while post-rolls have a 23 percent viewing rate. Of the pre-rolls being viewed, 30 percent are being viewed in full, whereas the others have some type of trick mode usage.
The opposite holds for post-rolls – 65 percent are viewed in full. Clearly, next-generation advertising systems need to support different handling (and possibly charging) mechanisms for advertisements played, skipped, telescoped (similar to the pay-per-click model) and interacted with (with different logic for pause, rewind and fast-forward).
One interesting trend that I found is that about 30 percent of viewers watch less than 5 seconds of an advertisement. This made me think of Comcast Media Center chief scientist Dan Holden’s presentation on dynamic ad skipping at The Cable Show ’09, where he suggested that viewers lose interest with irrelevant ads quickly, and therefore need a more relevant ad. Optimizing an ad slot such as this is something a video-on-demand ad system could easily support, as each interaction can be interpreted as another playout request.
While we’re all working to design the most flexible systems, we have to be aware that the advertising system itself is unable to support the sophistication that we may put into our systems. If we create the ability to support hierarchical charging with thousands of dependencies, our buy/sell side can’t implement the execution of that level of sophistication.
SO WHAT ARE WE TO DO?
Let’s watch the key trends and make sure our systems are able to remain flexible, yet support key market factors. For example, ad insertion technology should be able to understand and support ad priorities – these could be based on metadata, placement factors, CPM, etc. – and these priorities would interact with bandwidth allocation limits, storage limits and caching, as appropriate. Likewise, obvious flexibilities such as trick mode usage are proving their importance during these early trials.
Let’s build our systems flexible, yet realistic based on the data we’re seeing, and then we can start to move from the design and standards phase to the real implementation phase. I know many of you are with me on that one.