短視頻傳播的底層邏輯

The birth of short videos is a more intuitive and direct form of expression that has emerged after the era of text, images, audio, animation, and long videos, due to the development of basic equipment and network technology. It lacks a gradual process and requires something that directly stimulates the desires of the largest audience at the lowest level in order to gain dissemination and maximum traffic. The purpose of playing short videos is to quickly gain a large amount of exposure. Conversion is determined based on the usage scenarios of the product, which is not considered for now. The focus is on how to achieve precise exposure. This is also the theme of this article, how to create "high-quality content" that conforms to the underlying logic of short video dissemination through understanding the underlying logic of short video dissemination, and ultimately gain a huge amount of exposure.

The underlying logic of platform algorithms#

From the perspective of Douyin, it has 600 million daily active users, which means it needs to allocate at least half of these 600 million exposures to official activities to achieve operational goals. A portion is allocated to major events and hot topics of the day, and a portion is allocated to the promotion of large official institutional accounts (such as advertising and selling traffic). Of course, this is my personal speculation, and the specific allocation method is definitely a secret of Douyin. The remaining approximately 100 million traffic is distributed among 60 million creators. This is the current situation of traffic distribution, and each person does not receive much traffic, so an algorithm mechanism is needed to allocate it.

From the user's perspective, you are a consumer of Douyin. Douyin will judge your preferences and behaviors comprehensively, establish your personal profile model, and make comprehensive judgments on multiple dimensions to determine what content should be recommended to you in order to increase your activity and trigger purchasing behavior.
From the perspective of creators, your daily behavior of browsing Douyin and your creation traffic are completely unrelated. Douyin will make a basic judgment on your account based on your daily behavior to determine if it is a normal account. The rest depends on the quality of your content and the weight of your field.

The weight of the field is a prerequisite#

Douyin establishes a creator model for you based on the work data, looking at what kind of people are attracted to your account's continuous creation of content.
So your field is determined by this group of people, and your content will be continuously pushed to similar people. It does not rely on fixed tags, but uses a crowd profile model to determine the field. For example, for a painting account, if the creator receives likes from people who like painting every time, it will be determined that this group of people likes their content, and it will be recommended to similar people next time.
Even if the creator does not post paintings, as long as this group of people who like painting still like and comment on their content, it can still generate a large amount of traffic. The popularity and content are not directly related, but are related to the audience data generated by the content.

How to judge the quality of content?#

The algorithm will initially judge whether the content is normal, including whether it is original, clear, and related to previous content (randomly checking 30 images per second). It also checks if the content has participated in activities and if the music used is popular, and whether it is related to hot topics, etc. If the algorithm and machine consider the content quality to be good at this step, it will determine an initial distribution of traffic for you, and whether you can participate in the competition with the other 60 million creators.
The remaining step is how the specific group of people judges your content. If they have a high level of interaction with your content, Douyin will consider that this group of people with similar tags likes your content, and then recommend it to more people with similar tags. This is like a horse race, from the perspective of this group of similar users, they will be recommended to many contents that they may like, and those who provide feedback will stay, this is the horse race.
When the content receives over 100,000 likes, it will trigger manual review (because the level of dissemination is terrifying and requires stricter consideration of platform influence). The manual review will judge whether the content conforms to values and rules. If there are no issues, it will continue to be recommended until it becomes a popular topic with millions of views. This is the entire path of content dissemination.

What is a short video from the user's perspective?#

(1) Action 1: Seeing#

There are two possibilities in this step. One is the recommendation page, where you cannot see the cover, and the other is seeing the cover and title on the homepage.
When scrolling through the recommendation page
There is only 1-3 seconds to evaluate the value of the content. If it doesn't catch your attention, you will swipe away.
Actively viewing the homepage
Evaluate whether to click and watch based on the cover content, text title, and likes from others.

(2) Action 2: Evaluation#

Because it is a short video, there is only 1-2 seconds for evaluation, not enough time for thinking, and it is done subconsciously.
Form a simple expectation through comprehensive evaluation, whether the video is (useful, interesting).

When will the expectation of "this video is useful to me" be generated?#

Identification of identity group, this video is relevant to me, may be useful.
Identification of regional identity, this video is relevant to me, may be useful.
Directly asking questions at the beginning, I also have this question, may be useful.

When will the expectation of "this video may be interesting/funny/sad" be generated?#

Funny music rendering, this video may be funny.
Emotional music rendering, this video may be touching.
Curious scene, this video is different, may be interesting.
Beautiful women appearing, this video may be attractive.
Music for costume changes, this video may have a twist.
This expectation is formed by the combination of cover, copywriting, topic tags, music, and scenes.

(3) Action 3: Motivation for Interaction#

Forms of interaction include likes, comments, shares, and follows, and the motivations for each behavior are different.

Motivation for liking:#

Emotional value, feeling happy, joyful, evoking memories, sadness (teaching, pets, inspirational, love songs).
Self-commitment, hoping to shape a better self, an ideal state (good lifestyle, goals, etc.).
Collecting value, book lists, playlists, unfinished movies, content to be revisited (skills, checklists, appearance, stories).
Astonishing effect, unseen gameplay, unseen world, unexpected twists (satisfying curiosity).

Motivation for commenting:#

Information density: discovering details that others haven't noticed, so commenting (intentional flaws, small details).
Strong resonance: such as ethnic pride, moral judgment, such as "only fairies can see this video" or "you are amazing for finding this video".
Social interaction: hoping friends will also see it, usually being guided, such as "isn't this your friend?"
Unveiling mysteries: such as questions raised in the video, some people will look for answers in the comments, and naturally, others will answer.
Controversial topics: some opinions are easy to debate, such as the importance of education, gender topics, etc.

Vanity: establishing an advantageous position in social interactions, such as strategy tips, news, gossip.
Identity symbol: expressing one's own opinions, conforming to one's own identity image, such as looking down on others, self-indulgence.
Emotional resonance: driven by self-emotion, memories, pride, anger, etc., people are more likely to share.
Instant feedback: after sharing, you can immediately get something.

Motivation for following:#

Long-term interest: high appearance, good plot, satisfying curiosity, desire for expression, desire to peek.
Long-term usefulness: continuously sharing practical knowledge, demonstrating specific skills, satisfying the desire for knowledge, gaining a sense of achievement.
Common characteristics: series with numbered covers and titles.