Video viewers endure a hefty penalty these days just to watch a video; they need to persevere during lengthy loading times – prolonged for the purposes of loading the most valuable video advertisement – until they finally reach the desired content.
Is it any wonder why viewers are frustrated? How did we get here?
It all started with the old waterfall model. The waterfall model prioritizes which partners bid for an ad spot, beginning with direct buyer ads, and then moves on to the next line of network buyers, SSP partners or other exchanges. Unfortunately, not all partners can participate concurrently. This model’s daisy chain process becomes extremely time consuming, causing a drop in competition, CPMs and fill rates, as well as increased latency.
Even though publishers have been increasingly weary of users’ disinterest in their content due to these latency issues, they’ve been equally eager to maximize yield. Ultimately, this eagerness led them to fully embrace the new technique on the scene – header bidding, a programmatic approach that promised to increase available demand to all buyers simultaneously, democratizing the marketplace.
Yet just like the waterfall technique, today’s header bidding has serious issues. Especially for video advertising.
Standard header bidding is supposed to offer each impression to multiple demand sources simultaneously and increase yield by opening up the process to as many potential buyers as possible. This model appeals to buyers seeking to sidestep and cherry-pick inventory, and to publishers seeking access to all demand sources. Seems perfect, but in practice, publishers are loading their header-bidding with multiple concurrent real-time bids (RTB) of various sorts and sources, creating a complex setup process that requires additional resources.
The dysfunction of header bidding is being hotly debated and Google has sought to eliminate header bidding altogether by employing their own maneuvers (opening up dynamic allocation to outside demand). Still, with 70 percent of publishers adopting the model (cited from Business Insider), header bidding is not going away and remains the choice model for most publishers, including those utilizing video ads.
Though when it comes to video advertising, header bidding faces quite a few stumbling blocks, being that Video is very different than Display, making every piece of the puzzle more complex. Loading times are longer, different playback and ad technologies are managed in separate silos, and a myriad of different mobile devices and operating systems require specific adaptation. Additionally, latency matters only get worse for the new and increasingly popular Outstream Video format, where video content plays between paragraphs of text, reducing the effective window of opportunity to programmatically acquire ads from multiple sources.
Latter day versions tried to fix the flaws with a patched-up data-driven approach, inserting compromised customizable wrappers, mediation layers or extra code into the video player itself. Unfortunately, most of these optimization hacks end up “brute-piling” a set of parallel redundant calls that once again produced latency and content load delays, resulting in aggravated viewers.
Some industry players tried to reduce latency through “RTB style” server-side optimization. Unfortunately, these server-side-only tactics are plagued with their own downsides. They lack transparency, are highly vulnerable to fraud (especially on mobile), and reduce the scope of effective demand becoming available for publishers.
So now that we’re here, what’s the solution?
With video ad views growing at a 32 percent rate-over-year (IAB), publishers are increasingly aware of the need to find a better way to run and monetize video advertising and transform header bidding.
To run video advertising effectively, video ad formats (VPAID) require authentication on the client-side of the video player. However, standard header bidding implementations simply trigger multiple ads on the client side. This reduces the level of transparency to the demand side servers, thus causing redundant requests that slow loading times and compromise viewing experience even further.
For header bidding to truly work in video across all formats, In-stream, In-place, and Out-stream, and to overcome the barriers on both the client and server side, it needs to function in an open, two-sided environment.
It also needs to activate predictive optimization with ultra-precise timing, in a bid to manage an unlimited number of demand and supply sources in real time. The process of optimization needs to begin with heavy lifting on the server side before any bidding takes place. Then the “last mile” of yield optimization should be completed on the latency-sensitive client side.
So, on the client side, this model can predict video fill-rate optimization, prioritize the possibilities, and then make the necessary changes needed to produce effective video monetization with minimal latency impact. This approach generates lift and makes big improvements in overall yield while significantly reducing fraud and latency across mobile and other devices.
Header bidding in an open and two-sided environment would ensure transparency, resulting in a significant reduction of fraud related to source authentication, as well as offer a much-needed improvement in video loading speeds. All-in-all, drawing the best from both worlds on both the client and server sides, this model is in an extremely credible alternative to standard header bidding, overcoming the model’s present-day pitfalls.