An article explains the secrets of Xiaohongshu's recommendation algorithm

In-depth analysis of the operating mechanism of Xiaohongshu's search traffic, recommendation traffic and live broadcast traffic algorithms reveals the key factors for success on this platform. Let's explore how to better use this platform, from search to transaction, to achieve brand and personal value growth.

Every time you open the site, you start to plant seeds; every time you leave the site, you start to close a deal.

Xiaohongshu’s “shopping guide” attribute and decentralized recommendation mechanism determine that the usage path of most Xiaohongshu users is: search keywords → browse recommendation pages → make a deal

These three actions also correspond to the three user usage scenarios of discovery page, search page and live broadcast page respectively. The algorithm distribution process is different in different scenarios. If you want to make good advance arrangements, you need to understand the algorithm distribution logic of different scenarios.

Today, I will unveil the mystery of the algorithm and talk about the algorithm distribution logic of Xiaohongshu search traffic, recommended traffic and live broadcast traffic.

01 Search traffic algorithm

Based on Xiaohongshu’s platform positioning, more than 65% of its traffic comes from search, so the search traffic algorithm is more sophisticated. So here I will focus on the logic of the search traffic algorithm.

Publish a note - review - include - tag classification - first round of recommendation test - greater exposure. This is the algorithmic process that a note will go through from publishing to exposure.

Based on this algorithm logic, as long as your notes pass the review and get exposure, no matter how many small eyes you have, it is considered to be recommended by the algorithm.

First, the user searches, and then the system performs algorithmic matching based on the search terms and displays all the results. If the keyword is a relatively large term in the relevant category, then you can see some special label terms in the upper half of the interface that provide classification and filtering functions. This method will provide a better experience for users who search aimlessly. At the same time, the most popular notes are ranked first. The purpose of this search result display format and filtering conditions is to narrow the selection range and help users quickly choose.

The matching between search results and needs is mainly the matching degree between core keywords and queries. The specific content displayed in the search results is to find the information that best meets user needs by analyzing user needs.

The keywords in a note title are of utmost importance, and the official also clearly reminds you: "Fill in the title to get more likes"

It can be seen that the title is an important option used by Xiaohongshu to identify content attributes. If you want your notes to be more visible, the most basic work is to optimize the title.

We should make good use of search keywords, hot word recommendations, etc. to help us find the core words of the notes so that the system can recognize and recommend them to the corresponding users.

1. Find core words from recommended content

Recommended content includes several aspects, grayed-out keywords in the search box, historical searches displayed on the page, and hot search terms.

1) Default prompt word

Before you click on the search and enter the search term, the platform will recommend default prompt words based on the user's tags. There is a certain amount of search traffic in the default prompt words

2) Search discovery (hot searches)

Hot searches display the words that have been searched the most recently, to guide users to see some recent popular content and topics with high user search volume, which are related to the user's search volume and recent hot topics.

3) Supplement associative keywords

Supplementing associative keywords means that the user inputs part of the content, and then the system associates the complete content based on the content, automatically completes the keywords, and increases the user's choices by matching keywords instantly and displaying them. I searched for "slimming" and the platform recommended several keywords related to "slimming" to me.

Considering that the ranking of hot words is the result of comprehensive display, in addition to the number of notes, the ranking of the popularity of "hot words" may also involve two factors: the frequency of active searches by users, and the popularity of the notes themselves recommended by the system.

After searching, the system performs algorithm matching based on the search terms and displays all the results. If the keyword is a relatively large term in the relevant category, you will see some special label terms in the upper half of the interface that provide classification and filtering functions. This method will provide a better experience for users who search aimlessly. At the same time, the most popular notes are ranked first. The purpose of this search result display format and filtering conditions is to narrow the selection range and help users quickly choose.

There are a few things to note about keyword selection:

1. Xiaohongshu's hot search recommendations are indicators of the platform's short-term traffic content; search prompt keywords and popular screening are where long-term traffic is, and are derived from Xiaohongshu's real user data analysis and summary.
2. Be sure to give priority to keywords with low competition, high traffic and high accuracy, and avoid choosing broad keywords.
3. Learn to reverse deduce keywords. After determining the topic and keywords of the note, reverse deduce what keywords you hope users will use to search for your notes, and consider what common keywords you would use to search for this type of note.
4. Reasonable placement of keywords in the title, body, topic, and comments of the note will help the note to be included and accurately recommended. Avoid keyword stacking, which will be judged as advertising by the system. If you do this for a long time, your account will be downgraded by the system.

If we can eat up all the brand-related search traffic on Xiaohongshu, it will be the best solution for brand-effect integration.

02 Recommended traffic algorithm

Search is not enough. If you want your notes to get more recommendations, you also need to match the content of your notes with the content algorithm.

After a note is published, it will first be tagged with a series of tags by the system, and then recommended to fans who are interested in these tags. For example, if you like to read skin care articles, the platform will recommend more skin care articles to you.

After the note is pushed to the fans, it will be scored internally based on a series of interactive behaviors of the fans to decide whether to continue to push it to other fans. This is the old CES algorithm (community engagement score).

CES = 1 point for like + 1 point for favorite + 4 points for forward + 4 points for comment + 8 points for follow

This algorithm has actually been used in Xiaohongshu since 2017. Today, the algorithm's judgment indicators are richer and more diversified than before. Based on the practical experience of advertising in Xiaohongshu, we found that the platform now focuses more on the following items:

Click-through rate, interaction rate, completion rate, interaction value

But no matter how the data indicators change, the core of the content algorithm is still to examine the fan interaction behavior brought by the notes.

The interactive behavior of fans indirectly reflects the content quality of the notes. Even for experts with a small number of fans, if the scores are high, the notes will be recommended to more fans by the system, forming a "ladder" algorithm recommendation.

Notes with high scores will further receive traffic from searches on Xiaohongshu and Baidu. This type of traffic is very sustainable, and notes have a strong long-tail effect, which can maintain the growth of likes and comments even after being posted for several years.

Of course, in addition to the content itself, account weight is also important.

When your account has a certain number of followers, the notes will be recommended to your followers. If your followers like the notes, they will be further seen by their friends, forming a fission-like growth.

Xiaohongshu itself is an e-commerce platform. Many users come to Xiaohongshu with a similar mood of "shopping". It is fundamentally different from pan-entertainment platforms such as Douyin. Although it is difficult for traffic to grow "explosively overnight", it has a significant long-tail effect and a long exposure cycle, so the probability of fans being attracted is very high.

As long as the content is high-quality and the keywords are matched accurately, users can continue to be attracted to the content and attract social fission.

03 Live streaming traffic algorithm

Match the user's search needs, attract user attention through content, and finally reach the step of promoting user conversion

Xiaohongshu's e-commerce algorithm is similar to Douyin, allocating weight and traffic.

1. Weight determines the amount of streaming

What is weight? Simply put, weight is the evaluation made by the system on each account, and this evaluation is measured by comprehensive data. The higher the platform's rating of you, the higher the weight, and the more traffic you get.

In Xiaohongshu e-commerce, "weight" is mainly divided into: basic weight and real-time ranking weight

1) Basic weight

As mentioned above, the basic weight is different for each account. The system will determine the weight level based on the comprehensive performance of the account within a certain period, and it is not fixed. If you don’t make progress, you will regress. The fittest will survive.

But do you think that you can sit back and relax and sell products all the time just because your basic weight is high?

Don't forget, there is another weighting mechanism

2) Real-time ranking weight

Each account has a basic weight. The amount of traffic after the broadcast is determined by the weight level. However, the live broadcast room is like a reservoir. If there is no new traffic coming in, there will be no one in the live broadcast room soon. What is the distribution mechanism for subsequent traffic?

Xiaohongshu uses a real-time horse racing mechanism, which means that after you start broadcasting and get a wave of traffic, if you want more traffic in the future, you need to compete with your competitors at the same level.

The system is ranking every moment. 5min, 30min, 60min, Douyin live broadcast traffic is constantly running a horse racing mechanism to screen out high-quality live broadcast rooms and allocate more traffic to them.

(1) First, the initial push flow:

The initial traffic level is the basic weight we mentioned above. The higher the weight, the higher the initial traffic. At the same time, the word-of-mouth score will also affect our traffic.

(2) Enter data evaluation:

The system uses data to evaluate the performance of your live broadcast room every 5 minutes, 30 minutes, and 60 minutes, mainly based on interaction data and e-commerce data (specific key indicators will be discussed separately below)

(3) Performance on the best-selling list:

Then the system will sort and compete with each other at the same level. If you are higher than the previous one, you will enter the next level of traffic pool and get a new wave of push traffic. Then, the data evaluation will be repeated. If you are lower than the next one, the push traffic will be reduced, or stopped, or even returned to the initial traffic level.

According to this mechanism, real-time ranking is constantly carried out, and the fittest survive. Therefore, only by planning and executing every detail of the live broadcast can you beat your competitors and get more free traffic.

But traffic is not the goal, conversion is. The premise of pursuing conversion is traffic accuracy.

So, how does the system match us with precise groups of people?

2. Tags determine the quality of streaming

The tags here can actually be understood together with the content algorithm above.

In the e-commerce channel, the platform will also classify users according to different categories and group labels, match them with the content labels of the live broadcast room, and try to recommend them to interested users first, and then further refine the live broadcast room labels based on the effective viewing, stay, comment, like, conversion and other data indicators of interested users, so as to make more accurate investment.

In addition, corresponding to the weights above, tags are divided into tags under basic weights and real-time tags.

1) Basic weight label

Above we said that the basic weight determines the amount of traffic pushed by the system, so what is the relationship between the basic weight and the label?

The basic weight is formed by interest tags and e-commerce tags. To tag the live broadcast room with interest tags, you only need to design the people and goods field in the live broadcast room and plan the script to attract target users to watch the broadcast, stay, interact, and become fans.

E-commerce tags require the accumulation of historical e-commerce orders to give the account an accurate e-commerce tag. By doing a high-density transaction over a period of time at the beginning, the account can be labeled with a basic e-commerce crowd tag.

2) Real-time tags

This is easy to understand. Real-time traffic forms real-time tags. In each live broadcast, we need to use precise product planning and paid traffic to continuously deepen account tags. The platform will explore the interactive and transaction groups in real time, and the push model will become more and more accurate.