Relevant content
If we are talking about what is the main key to the success of any content project, the answer will always be the same — to show people what they are interested in. If people like what they see, they share it with friends and come back again and again, providing an increase in the number of active users.
This is easy to do when we talk about standard social networks, such as Facebook or Instagram, where user interests are built around their social graph. We follow the people we know or who are interesting to us, to know what is new and what they think about. Communicating with friends on the Internet has become an integral part of our life and we can not get out of it.
It is also simple when we work with one narrow audience that has similar interests, and all content is created or selected by internal editors taking into account the interests of a given audience. For example, if we are interested in start-ups and technologies, we feel free to go to Techcrunch, where we can always find the latest news from the world of technology or interesting articles on this topic. Thus, our expectations clearly coincide with the content that we find on this specialized resource, because its editors know exactly what we want to find, and monitor the quality and direction of the content.
The problem arises when we start working with a wide audience, and the users themselves participate in the creation of content. In the absence of a given editorial policy, the content of different users may be very different, moreover, each of them may have their own, and completely different, interests, their own style and their own ideas about quality.
General feed
Of course, the problem of displaying the most interesting content can be solved by the standard hot feed algorithm, which considers the rating of the post, based on the time of its publication and the number of votes received, thereby raising the content that users rated the most, and over time lowers it to give way to new interesting posts.
In general, such a system works quite well. Its plus is that any user, regardless of how many subscribers he has, can quickly go up and thousands of other people will see his content if he has published something really interesting.
However, such a system has one big disadvantage — it takes into account only the opinion of the majority. Trying to make the feed interesting for everyone, we at the same time make it not very interesting for each individual user. The fact is that the preferences and interests of each individual can vary greatly. Moreover, the narrower each selected subject, the more loyal its fans.
Within the framework of a common feed, authors who create content on a specific topic will almost never be able to get enough attention, because a priori they will lose the war to more general content. Such authors lose their motivation to create narrowly focused content, as it is difficult for them to find their audience, and users who might be interested in this topic cannot find what they were looking for.
Smart feed
The solution to this problem is to create a smart feed algorithm, which, by means of machine learning and analyzing the votes, subscriptions, user views, and etc., will offer users the most relevant content. However, in order to offer users content that reflects their interests, you must first create the conditions for the appearance of this content. We understood that we had to start creating such conditions if we wanted to become a truly competitive media, and that is why we created content categories.
Content categories
Content categories allow us to divide content in the app into 11 general areas: Memes, News, Video, Aww, MDK, Lifestyle, Discussions, Science, Crypto, Gaming and Art. Now, when creating a post, the author must select one of the content categories, and other users, in turn, in the settings of their feed will be able to choose which categories they are interested in and watch the content only by them. This allows, firstly, to take into account the interests of the audience and offer them the content in which they are most interested, secondly, to motivate authors to develop other, narrower directions, because now their posts will be easier to find and they will not get lost in the general feed.
Adding content categories is only the first step towards a variety of content and the formation of a relevant personalized feed. The fact is that even within the framework of one category there can be many directions divided by narrower interests. For example, memes can easily be divided into black humor, text jokes, memes about relationships, or even just screenshots from the Tinder.
Next steps
The next stage in the development of personalization in MDK will be the creation of moderated tags. It is assumed that users will be able to create their moderated tags and set the theme of these tags. Other authors will be able to post to these tags, and moderators will decide whether this post is suitable for their topics or not. So, such tags can form their own separate audience, and top posts from these tags will be strained into the feed of those users who are subscribed to them, providing them with relevant views. Now we are working on this idea and will definitely tell about it a bit later.
Stay tuned to watch as we change the social media industry.
