Latency certainly matters for video viewing…

Two scientists from Akamai recently published a paper entitled Video Stream Quality Impacts Viewer Behavior: Inferring Causality Using Quasi-Experimental Designs. Now Akamai being in the business of optimizing internet traffic flows and particularly internet video flows, you’d expect the paper to conclude that low quality in video streams causes users to disconnect, and indeed it does. Still, if you accept the inherent bias that Akamai can only measure videos they’re enabling (and therefore presumably not Youtube or Dailymotion), the methodology seems sound. Here are the main results as quoted from the abstract:

We study the impact of video stream quality on viewer behavior in a scientific data-driven manner by using extensive traces from Akamai’s streaming network that include 23 million views from 6.7 million unique viewers. We show that viewers start to abandon a video if it takes more than 2 seconds to start up, with each incremental delay of 1 second resulting in a 5.8% increase in the abandonment rate. Further, we show that a moderate amount of interruptions can decrease the average play time of a viewer by a significant amount. A viewer who experiences a rebuffer delay equal to 1% of the video duration plays 5% less of the video in comparison to a similar viewer who experienced no rebuffering. Finally, we show that a viewer who experienced failure is 2.32% less likely to revisit the same site within a week than a similar viewer who did not experience a failure.

Emphasis mine. Now this is nothing new, I think everyone is aware of that on some level. What astounded me was how little tolerance users have for non-delivery of a video. The graph provided in the paper (see above) is quite telling also: it shows not only how fast customers drop off, but the fact that the better connectivity they’re using, the less tolerant they are to latency in getting the content they’re trying to view. Fiber users are the least patient of all.

This leads me to three quick (and hopefully interesting points):

  • first, as I’ve mentioned before, speed is addictive and users who get used to the comfort of low latency clearly lack patience when content delivery won’t follow,
  • secondly, as highlighted in Numerama yesterday (quoting the same study, in French), service providers who deliberately throttle video and/or specific video content providers, like Free currently does in France with Youtube are damaging the business of these content players, but also irritating customers. There’s only two ways this can go: either customers don’t care enough to churn and ultimately the Youtubes of this world will have to cave in and fork out to get quality delivery, or the customers care enough and leave visibly enough to force the service provider to actually provision the service properly. I know I’m one of the latter, and the first service provider who will provide me with a good offer from now on I’ll switch to. I’m a heavy Youtube user, and getting decent service on that matters more to me than a reputedly lame blueray player I’ve never used in my set-top box,
  • finally, I find it very interesting that the tolerance towards mobile is so high. Clearly, there’s no alternative if you’re viewing on a mobile (if you were in a wifi enabled area, you’d be using wifi and it would count as whatever the wireline technology is in the statistics), but I still find it amazing that customers are willing to wait so long for a video on mobile. I’m wondering how long that tolerance will last…