Eliminating the Ad-Related Culprits in Video Latency
Viewers hate waiting... But does waiting for content to load actually impact user engagement? Publishers instinctively know that better user-experience leads to a larger, more engaged audience, which in turn means more revenue. To help quantify this correlation, Akamai conducted a study (PDF) to measure the correlation between online video load delays and user drop-off rates. The results suggest that user patience may be even lower than intuition might suggest. After just two seconds, 20% of viewers give up and leave, after each additional second a further 6% leave.
Needless to say, losing a quarter of your revenue opportunity because your video player is slow creates a significant incentive for publishers to improve!
The study focused on latency caused by poor bandwidth or network connection speeds. But in many cases video load delays are actually traced back to pre-roll ad calls. Clearly, reducing any delays created by advertising can dramatically improve user experience and retention, increase inventory and increase revenue. Now, more than ever, it’s critical for publishers to optimize their video ad infrastructure to help reduce these latencies.
Fingering the Culprits
There are several factors that contribute to increased advertising-related latencies, but the most important are: ad server infrastructure, CDN performance and interactive ad units.
Minimizing ad server infrastructure latencies can be a challenging task and more often than not it is about building an architecture that scales – efficient ad serving algorithms, built on fast, efficient languages like Java and C. Geographical distribution and optimal routing, robust capacity monitoring and planning procedures also help.
The actual rendering of the ad requires the video player to load the ad creative asset from a remote location, usually a CDN endpoint; it is therefore critical to consider CDN performance as an important factor and select a vendor that operates a vast infrastructure of geographically distributed low-latency edge servers.
If the video ad creative is a flat media file, then CDN performance alone can guarantee minimal delays on ad rendering. But if the ad asset is interactive (VPAID), then latency becomes an unknown because it is no longer under the control of the primary ad server. A third party interactive ad unit has its own loading and rendering cycle, which can dramatically increase the delay of the main user experience. If advertising-related latencies are to be minimized, it is recommended to avoid serving interactive ad units, or to constantly audit these units and quarantine high-latency assets.
With these key performance-hindering factors mitigated, load delays can be significantly reduced and the entire cycle, from user initiation to successful ad rendering, can take just a few hundred milliseconds. Compare this to as much as three to four seconds to serve an ad from a tier-2 ad platforms, and its clear that using the right ad infrastructure can help publishers retain as much as a third of their audience.... and their revenue.
Andrei Dunca co-founded video ad platform LiveRail with Mark Trefgarne and, in his role as Chief Technology Officer, coordinates the design and implementation of LiveRail's technology stack, products and infrastructure. Prior to joining LiveRail, Andrei co-founded and was the CTO of Trilulilu, the highest ranking Romanian website. Andrei is a Computer Science graduate of the Technical University of Cluj-Napoca, Romania, and is currently based in San Francisco.