Toutiao’s personalized recommendation has become popular with Toutiao and Douyin. What is the magic of the recommendation and recommendation system hidden behind the explosion?
(1) Concept of recommendation
Based on massive data mining, recommendation provides users with personalized information services and decision support. Recommendations can be classified into personalized recommendations and non-personalized recommendations. Recommendation has three functions: reduce information overload, discover the long tail, and improve conversion rate. It is worth emphasizing that personalized recommendation is only one kind of recommendation, which is accurate to a person. Non-personalized recommendation includes popular recommendation, editor selection, similar recommendation, etc.
(2) The concept and core value of recommendation system
As the name implies, recommendation system is a mechanism for sorting and filtering information reasonably according to the differences of users and scenarios to solve the problem of information overload. Recommendation system helps users find the service and information they need in the mass information, which is the product of the information explosion era. Since the reform and opening up, material and cultural life has been greatly enriched, and the threshold of content and commodity production has been constantly lowered. Faced with information overload and limited energy, the speed of information acceptance is in contradiction with the speed of content and commodity production. In this case, the core value of the popularization and application of recommendation system is to solve information overload. In brief, we can understand the recommendation system from the following four key points, as follows:
Key 1: The difference between users and different scenes is a point that many enterprises tend to ignore in the initial stage of making recommendation systems. Recommendation systems are not only based on users, but also the difference between relevant scenes will have a huge impact on the final recommendation effect.
Key 2: Reasonable sorting and filtering of information In fact, the enterprise has tens of thousands or even hundreds of millions of items, which may be articles, videos, goods, etc. How to push these items to users will involve the operation principle behind the recommendation system.
Key 3: Recommendation system to solve the problem of information overload Enterprises need to help users solve the problem of information overload, so as to design such a mechanism for users.
Key 4: A set of mechanisms A recommendation system is a set of mechanisms composed of different algorithms, rules, etc.
(3) The value of recommendation system to enterprise products and business
The value of recommendation system to enterprise products and business is mainly reflected in the following aspects: First, recommendation system improves product intelligence and user experience. Taking the content distribution industry as an example, editors in the portal era distribute content within channels by manual means.
In the case of a sports channel, the granularity of the content an editor focuses on May be the NBA, but the granularity of the user base that follows “James” is not well satisfied by the editor. Second, the recommendation system reduces operating costs and improves operating efficiency. With the disappearance of flow bonus, enterprises focus on cost reduction and efficiency increase. Traditional portals may require hundreds or thousands of edits to distribute content and still cannot meet users’ more granular content needs. The existence of the recommendation system can realize that several operators are responsible for the distribution of the contents in a client and do some recommendation intervention on a daily basis.
Want to learn more about recommendations? Click on the link to download the complete white paper on building recommendation Systems from Zero to one
White paper on Building recommendation System from zero to one