A new study from Akamai released today offers more insight into the importance of video quality on viewer engagement and brand perception. The study also found that the business model used by the streaming provider influenced the extent to which quality influenced perception of the video experience. Titled "Bit Rate and Business Model," the study used new technologies to evaluate viewers' reactions to video streams at different QoE levels.
Most people in the industry I have discussed this with agree that QoE is critical for streaming providers. Various surveys have also identified it as important, though in a prior survey conducted by Heavy Reading, it was specifically buffering and rebuffering that were anathema to viewers, not minor fluctuations in video quality (for example, switching to a lower profile in an ABR scenario.)
Still, as Akamai pointed out in its report, what people say in surveys is not always what they do. The CDN provider offered an example of a recent TiVo survey where 18% of respondents said they would cancel their pay-TV subscription within six months, and a further 32% said they might. But in fact, only 9% actually ended up cancelling in that time period.
So Akamai turned to Sensum, a company that measures emotions using advanced technology: It uses facial coding, heart rate monitoring, implicit association testing and eye tracking as well as wearable technology such as galvanic skin response monitors.
The Akamai study included 1,200 respondents, divided into two groups. One was shown a high-quality stream with no buffering while the other saw the same video at a lower bit-rate and had a "re-buffering event" during a key scene. Each subject was also "primed to associate their own viewing experience with one of the three major OTT business models: Subscription Video on Demand (SVoD), Ad-Supported Video on Demand (AVoD), or Transaction Video on Demand (TVoD)."
When viewing video uninterrupted by buffering, viewers were 10.4% more engaged emotionally with higher-resolution video than with lower resolution. When buffering was introduced into the lower-resolution stream, the difference in emotional engagement nearly doubled, to 19.8%. Interestingly, viewers using an SVoD service were 175% more responsive to the higher-quality stream than those using other models (AVoD or TVoD).
"Emotional" response was measured using galvanic skin response, which is based on the theory that skin resistance is affected by the state of sweat glands in the skin. The nervous system controls sweating, so skin conductance is a good indicator of psychological or physiological responses. Small monitors are attached to the fingers of study participants and can accurately measure skin conductance, and through it, viewer engagement.
The study also found that buffering caused a 16% bump in negative emotions. This was measured using facial coding, a technique used for more than 40 years to measure "momentary changes in facial expression that reveal emotions such as disgust, happiness, sadness, surprise, attention, and focus" according to the report. Buffering caused a 14% decrease in happiness, 9% increase in disgust, a 7% increase in sadness and an 8% decrease in focus.
Sensum also used implicit association testing, an approach measuring "gut" reactions and aimed at eliminating self-censorship. It asks respondents to rapidly characterize a concept or keyword without giving them time to think about a "perceived correct answer." Positive keywords were 20-22% lower for viewers of lower quality streams, with TVoD users particularly affected. "Captivating" scored 22% lower and "delighted" 20% lower among these viewers. Negative keywords also jumped for lower-quality streams, with AVoD viewers 43% more likely and TVoD viewers 23% more likely to rate the video "boring."
Sensum also used an old-fashioned survey at the end of the process, and found that three-quarters of viewers were at risk of churning if buffering kept occurring.
Frankly the new methodologies are probably more interesting than the findings. In a sense we have always known that QoE is important for video delivery, so this study doesn't change that perception. However, it does offer a measurable way to track viewer responses which could be more valuable over time. Of course all providers want to deliver the highest possible quality and have the happiest possible customers. But being able to track and quantify a "churn threshold" could be an extremely important piece of business intelligence.
For example, if you know that 45% decrease in emotional engagement is where you start to lose customers, you also know that a 40% decrease is then acceptable. That allows providers to optimize their investment in content delivery without losing users.
Of course, that is if these new methods are reliable over time and can be accurately used to track user behavior and satisfaction at a new, deeper level of granularity. For me at least, this is all very new and I hope to get more insight into these emerging research methods moving forward.
— Aditya Kishore, Practice Leader, Video Transformation, Telco Transformation