Relevance increases video consumption and engagement
Written by: David Maher Roberts
For the last several years, the online video world has focused on solving the challenges of how to get content digitized and distributed. But there’s been a lack of focus on figuring out how to increase consumer demand for your video.
Making every single video in a catalogue available is a good start, but that by itself will not increase the likelihood that your audience will watch more of your videos.
The question is: How to digitize as much of your catalogue as possible (i.e., supply) and offer individual users the content they are most likely to enjoy watching at that right moment in time (i.e., demand)?
The answer: It’s all about relevance. The more relevant the content choices (recommendations, charts, navigation, search) the more likely the user will engage and consume.
The proof that relevant recommendations drive consumption and engagement: Two-thirds of Netflix movie rentals and 35% of Amazon sales are driven by recommendations, and even YouTube has been talking about the need for smarter recommendations to drive consumption. My experience at The Filter is that online video publishers see an immediate and significant lift in video consumption, dwell time, and return visits when they replace editorial recommendations or popularity lists with more relevant recommendations using our algorithms.
How to make content more relevant to me, here and now
The key to relevance is data. Every single video-related service has two types of data: (1) catalogue-related data (description, metadata) and (2) consumption-related data (viewed, shared, rated, saved, commented). The more data there is, the better, but even modest amounts of data will deliver more relevant content to your consumers.
Capturing and logging catalogue and consumption data in a meaningful way is one of the best investments any video service can make today, because it makes relevance and personalization easier in the future. Luckily, data about consumers’ ever-changing preferences in video has never been easier to capture and understand. And more importantly, it’s never been easier to use this data to increase engagement, traffic, and revenue.
Sure, there is a lot of data and you need to use clever algorithms and have an acute understanding of how to use the math and data to deliver the best possible results. But there are companies like ours that focus on this every day and have built solutions for you to choose from.
Recommendation services like The Filter offer a lightweight and simple beaconing service to capture and analyze every bit of interaction with the content. Once this data is properly identified and logged, it can then be used by a recommendation and relevance engine. Using A.I techniques, the engine grinds through the catalogue and consumption data, and then works out the connections between every single piece of content in the catalogue based on individual audience members’ behaviour and consumption. This mathematical model delivers a list of relevant recommendations that go together in order of most-likely match based on consumption data from all users.
For example, if you watch a video of a kitten falling off a sink (yes – aren’t they funny…) on YouTube, the list of relevant videos that sits on the right would typically have lots of other kitten-related videos. This is because in simple terms, this list is a popularity chart using keywords as a filter. This type of recommendation leads to “kitten fatigue,” as there are only so many kitten videos you can watch in one session. Using the relevance model based on consumption and behaviour, however, the list of related videos includes those that have the highest probability of being watched after the kitten video, thereby offering more relevant rather than just similar content.
So presenting lists of recommended content that are more relevant to the consumer is better than an editorial- or popularity-based approach, but it is just the first step. Recommended content can be filtered to match the end user’s tastes and activities, can remove videos that he/she has already watched, or suggest video topics that he/she has a propensity to watch.
A fully personalized video service that takes into account tastes, past activity, and context is seen as the nirvana of relevance right now. It is very achievable as long as video services take the following steps to get there:
1. Identify and log catalogue and all types of consumption data as soon as possible
2. Plug in a proven and robust recommendation and relevance service and deliver more relevant content recommendations
3. Measure the increase in video views by individual consumers
4. Filter the list to offer personalized recommendations
5. Measure the increase again!
David Maher Roberts is CEO of The Filter, a recommendations & relevance platform for digital media content. The Filter is backed by music legend Peter Gabriel.
Comments