One way to increase the lifetime value of a publisher's video-viewing customer is to improve his/her user experience and make them a loyal, repeat customer.
To build customer loyalty, many publishers are turning to personalization, which creates more (and different) views by surfacing the most relevant videos in the publisher's library that most closely match and reflect user interests. It then programs them in a sequence that reduces bounce rates and entices more in-session and repeat viewing. This of course unlocks significant value as the publisher gets to run more high-CPM video ads.
The challenge however is to achieve personalization at scale. Human curation of playlists, which—similar to traditional linear TV programming—has a strong focus on editorial selection and quality standards but is a one-to-many recommendation that is not targeted to the individual user, cannot be done at significant scale.
Related: It Helps If the Audience Can Find Video They Like
Another programming method is automated selection. While this successfully resolves issues around scale, content selection is often limited to narrow parameters. Selection can be done at random or based on the characteristics of the previous video shown, a semantic metadata-based analysis and, at the moment, the most common method of content recommendation.
The primary limitation of these methods is that they confuse related and recent for personalization. Content publishers are not competing with other publishers for the user’s attention but every form of digital entertainment. Users enter a site or app for a variety of reasons and a variety of places. Getting users at scale to engage beyond the initial click requires automated systems to factor in a wide variety of circumstances. The ability to makes these decisions accurately requires analysis of data performed in real time. Poor content discovery impacts user experience and results in high bounce and low retention rates.
Furthermore, solely relying on programming mechanisms like playlisting, verticalized “channels” and social trending assets, have limitations and often impede rather than improve user experiences, consumption and monetization.
To remedy this disconnect between machine and human curation are proprietary personalization engines that use adaptive machine learning, artificial intelligence, and prescriptive business intelligence to program the content to viewers on a one-to-one level across all devices in real-time.
IRIS.TV's three dimensions of predictive analysis occur in real time while the viewer is watching. The technology seamlessly integrates into nearly every video player used by web and OTT publishers.
IRIS.TV uses artificial intelligence and adaptive machine learning to surface the most relevant video from the publisher's library for each individual site visitor. Before, users were watching a collection of individual videos. With personalized video programming, they tend to watch far more video and for longer periods of time. Personalized video programming has enabled publishers to increase video views by 70% on average. And it works no matter if you are publishing video on a desktop, tablet, mobile phone, or OTT.
Engagement is measured as views per viewing session, while retention is assessed through bounce rates and returning viewers.
In our own experience after the first month of using a video programming platform, lifestyle, entertainment, news, and sports publishers, on average, see a 70% increase in total monthly views, a 62% increase in user engagement in the form of views per viewing session, a 10% reduction in bounce rates and a 70% increase in returning viewers.
A metric that is very similar for all four publisher types is the retention of users over the course of the in-stream video programmed to them. Retention rates increase as a result of increased personalization. Most publisher types follow a similar path to retention. News, on the other hand, follows a more volatile path. This can be attributed to the late breaking nature of news. For instance, if the President gives a speech, it is likely the publisher would livestream or include the whole video in the initial position. This can result in a variety of experiences early in the stream, but at scale, the personalization curve resembles the path of typical shape.
We have seen that a portion of around 20% of highly engaged super-viewers watch four to eight videos on average. The value of those users who want to engage can be maximized with the provision of personalized streams not only with regard to videos viewed but also by programming branded video to them in-stream versus pre-roll ads. From both the publishers’ and the users’ point of view, personalized video programming is a win-win, as it creates value at both sides of the consumption process.
