Why is Douyin good? Not because of the four words “algorithmic recommendation”, but how it uses algorithms to reconstruct content distribution, user relationships and business models. In the second part of this series, we will dismantle Douyin’s key design in terms of content mechanism, user experience and growth strategy from a product perspective, and reveal how it has evolved from an entertainment tool to a super traffic entrance step by step. If you understand Douyin, you may understand the product logic of the second half of the mobile Internet.
In 2024, Douyin (including TikTok) will exceed 2 billion monthly active users, with an average daily usage time of 142 minutes and advertising revenue of more than 200 billion yuan – behind this set of data is an algorithm-driven attention revolution. When we disassemble the product evolution history of Douyin, we will find that it is not only a short video platform, but also a super application that uses technology to reconstruct user behavior, reshape business logic with content, and rewrite the Internet landscape with ecology.
This article will reveal the underlying growth code of Douyin from the three-dimensional perspective of product design, competitive strategy, and ecological expansion, combined with in-depth comparison with Kuaishou, Channels, and Bilibili.
1. Core Engine: The algorithmic revolution that redefines content distribution
1. Accuracy of the algorithm: from guessing what you like to knowing you better than you do
Douyin’s recommendation algorithm is known as the brain of the Internet that understands users best, and its core is not simply the user’s repeated push after liking, but through multi-dimensional behavior modeling + real-time feedback iteration, to build the user’s digital personality.
Technical details:：
- Explicit data: Likes, comments, shares, and completion rates (weight accounts for 40%) – directly reflect users’ preferences for content;
- Implicit data: Sliding speed (drawing away within 0.5 seconds→strong negative feedback), dwell time (more than 3 seconds→ potential interest), device environment (using mobile phones at night to push relaxing content→) (weight accounts for 30%) – capture the user’s unstated needs;
- Time dimension: Users brush the workplace dry goods in the morning, visit the store in the afternoon, and watch film and television commentary in the evening, and the algorithm will dynamically adjust the recommendation strategy (weight accounts for 20%);
- Social data: Videos liked by friends will be recommended first (10% weight) – The new social recommendation module in 2023 will increase the average daily interaction of users by 25%.
Compare Kuaishou: Kuaishou’s inclusive algorithm emphasizes that every user has the opportunity to be seen, but leads to large fluctuations in content quality (Kuaishou TOP100 video completion rate is 58%, Douyin is 72%); Douyin’s precise algorithm allows high-quality content to get more traffic through content screening + user matching (10% of Douyin’s top content contributes 60% of the playback, which is 1.5 times more efficient than Kuaishou).
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
User perception: When new users open Douyin, the completion rate of the first three videos is as high as 78% (industry average 52%), and the probability of the first use time exceeding 10 minutes is 63% – this kind of experience that cannot be stopped after a swipe is essentially the algorithm’s accurate sniping of user needs.
2. The evolution of algorithms: from collaborative filtering to multimodal learning
Douyin’s algorithm is not static, but continues to evolve with technology and user behavior:
- Phase 1.0 (2016-2018): Based on collaborative filtering, through the basic label matching recommendation of user-content (such as users watching beauty videos and promoting similar content);
- Phase 2.0 (2019-2021): Introduce deep learning to analyze the visual features (such as color, character movement), auditory characteristics (BGM rhythm), and text features (subtitle keywords) of the video to achieve content understanding (such as identifying that the core of the hot pot video is steaming ingredients rather than the restaurant environment);
- Phase 3.0 (2022-present): Upgrade to a multi-modal large model, integrating the user’s historical behavior + real-time scene + cross-platform data (such as the user has just searched for a dress on Taobao, Douyin will push a dress dressing tutorial).
Technical barriers: Douyin’s algorithm training data is 3 times that of Kuaishou, and the model iteration speed is 2 times that of the industry average. This positive cycle of data-algorithm-product makes Douyin’s recommendation efficiency always ahead of competing products.
2. Content ecology: the balance between industrial production and personalized consumption
1. Creation side: use tools + templates to realize the industrialization of national content
The content explosion of Douyin is essentially the result of lowering the creative threshold + guiding the standardization of content:
1) Lightweight tools：
- The shooting interface defaults to 15 seconds, and it takes an average of 8 minutes for users to complete a video (45 minutes for professional editing software);
- Built-in template shooting function (such as card point video disguise special effects), users only need to replace the material and generate professional-grade content in 30 seconds (template videos account for 41%, contributing 58% of playback);
- The clipping APP provides a one-click film creation function, allowing users who do not know how to edit to produce content (the monthly active users of Jianying exceed 300 million, covering 80% of Douyin creators).
2) Content structuring: Douyin challenges topical music lists through hot lists to guide content standardization. For example, in the 2023 Dopamine Dressing Challenge, the platform provides unified BGM, tags, and templates, and users only need to show their outfits to participate. This structured creation makes content more consumable – users are more likely to feel familiar + engaged when they see similar content.
Compare Bilibili: Bilibili emphasizes creative freedom, but the creative threshold is high (the average video duration is 10 minutes, scripting, editing, and post-production are required), and monthly active creators account for only 3% of users (Douyin is 18%). Douyin uses tools + templates to turn creation into participation, realizing a positive cycle of content production, distribution, and consumption.
2. Consumer side: Use content layering to meet the needs of all scenarios
Douyin’s content pool is not disordered, but covers the user’s full-scenario needs through hierarchical operation of explosive-long-tail-vertical categories:
- Popular content (accounting for 10%): Through hot topics + head experts, the goal is to attract users to stay (such as digging and digging children’s song videos in 2023, with more than 5 billion views, driving the average number of daily opens of users to increase by 15%);
- Long-tail content (60%): Produced by mid-waist creators, covering interest segments (such as cat raising daily notebook tutorials), with the goal of enhancing user stickiness (38% of users are retained by focusing on niche verticals);
- Vertical content (30%): Introduce professional content through industry support programs (such as knowledge partner cultural tourism promotion), with the goal of enhancing users’ sense of value (the number of knowledge video playbacks increased by 200% year-on-year, and the average daily learning time of users is 12 minutes).
Compare video numbers: Channels relies on social distribution, and the quality of content is limited by the number of interactions with friends (the average daily interaction of Channels users in 2023 is 0.3 times, compared to 1.2 times on Douyin); Douyin uses algorithm screening + hierarchical operation to ensure that the content pool has both explosive traffic and in-depth value, meeting the diverse needs of users for entertainment + learning + social interaction.
3. Ecological expansion: the path from content platform to super application
1. E-commerce: Reconstruct the relationship between people and goods with interest recommendations
When Douyin e-commerce was launched in 2020, Taobao (shelf e-commerce), Pinduoduo (low-price grouping), and Kuaishou (old iron with goods) had occupied 90% of the e-commerce market. Douyin’s breakthrough point is interest e-commerce – through algorithm recommendations, users can buy what they like, rather than searching for it if they need it.
Key Strategies：
- Content is strongly bound to products: Directly insert the shopping cart in the video (click to jump to the product page), and the anchor explains while showing during the live broadcast (the conversion rate is 2.3 times higher than that of the graphic detail page);
- Mall completion active search: In 2022, Douyin Mall will be launched, covering search + shelf scenarios, expanding from interest recommendation to active purchase (mall GMV will account for 35% in 2023);
- Closed-loop logistics and payments: In 2023, it will launch Yinzunda Logistics (in cooperation with Zhongtong and YTO, the delivery time will be increased by 30%), and Douyin payment will be tested (more than 300 million users are bound, and the payment conversion rate is 18% higher than that of third-party payment).
Data comparison: In 2023, Douyin’s e-commerce GMV will be 2.3 trillion yuan (Taobao 12 trillion yuan, Pinduoduo 3 trillion yuan), but impulse consumption accounts for 58% (Taobao 21%) – this shows that Douyin uses content to stimulate demand and open up an incremental market for e-commerce.
2. Local life: Use content to plant grass to break traffic + push barriers
When Douyin Local Life was launched in 2021, Meituan (in-store + takeaway) and Dianping (grass planting + trading) had established deep barriers. Douyin’s strategy is content-driven trading:
- Merchant contentization: Encourage merchants to shoot videos of the store environment (such as the process of freshly cut beef in a hot pot restaurant), and users can directly click on the group purchase coupon after swiping the video (merchants’ self-broadcast GMV accounts for 42%, and the conversion rate is 2.8 times that of traditional group buying pages);
- Experts bring goods: Mid-waist experts (10,000-1 million fans) released a store exploration evaluation video (such as 38 yuan to eat a double set meal, how does it taste?) ), commission sharing (talents take 10%-30%) to incentivize content production (the GMV of talents accounts for 51%, and the repurchase rate is 15% higher than that of Meituan);
- Instant delivery and completion: In 2023, it will cooperate with SF Express and Flash Delivery to launch hourly delivery, covering catering, fresh food, and daily necessities.
data: In 2023, Douyin’s local life will be GMV2200 billion yuan (Meituan’s 1.4 trillion yuan), but 71% of users place orders through video planting (Meituan’s 28%) – content has become Douyin’s nuclear weapon to break through local life.
3. Social: Implicit layout from content interaction to relationship precipitation
Douyin’s social networking has always been regarded as a shortcoming, but it has gradually built a relationship chain through implicit socialization:
- Weak relationship in the comment area: Users comment under the video and may interact with the publisher or other commenters (the average daily interaction in the comment area is 1.2 billion, 15% is converted into private messages);
- Relationships in the fan base: In 2022, a fan group will be launched, anchors will create a community, and users will pay or join interactively (the monthly active fan group is 230 million, and the average daily message in the group is 58);
- Shoot the same strong relationship: When users shoot the same video, they automatically @ the original author (32% of the same video contains @ original author, and 8% is converted into interconnection).
target: Douyin hopes to transform content interaction into social relationships, and eventually form a closed loop of content-social-consumption. Although social attributes are still weaker than WeChat, implicit social has become the key to user retention.
4. Competitive barriers: the triple moat of time, data and ecology
1. Time barrier: irreversible inertia of user behavior
Douyin users spend an average of 142 minutes per day, accounting for 28% of the total mobile phone usage time (industry average 15%). This time occupation forms a strong behavioral inertia – the user’s action of opening Douyin has changed from an active choice to an unconscious habit (neuroscience studies have shown that after 66 days of repeated use, the behavior solidifies into an automatic response).
2. Data barriers: full-dimensional precipitation of user-content-scene
Douyin has accumulated the world’s largest database of short video behavior:
- User behavior data: 10 billion interactions (likes, comments, swipes, etc.) are recorded every day;
- Content data: 1 billion videos are processed per day (covering visual, auditory, and text multimodal information);
- Scene data: Scene information such as the user’s time, location, and device.
These data form the fuel of the algorithm, allowing Douyin’s recommendation efficiency to continue to outperform competing products (Kuaishou’s data volume is only 1/3 of Douyin, and the recommendation accuracy is 20% lower).
3. Ecological barriers: the synergy of content + e-commerce + local life + social networking
Douyin’s ecology is not a simple function superposition, but a collaborative system in which content drives other businesses and feeds back content:
- Merchants shoot product videos: E-commerce provides product material for content
- Store exploration video: Local life provides scene material for content
- Friend Recommendation Videos: Social provides the power to spread content
This ecological collaboration has made Douyin’s content pool richer and richer, and users’ usage scenarios more and more diverse, forming a stronger and stronger Matthew effect.
Conclusion: Douyin’s product philosophy and the new paradigm of the Internet
The essence of Douyin’s success is the victory of the product philosophy of understanding human nature with technology, meeting needs with content, and creating value with ecology. It does not disrupt any industry, but connects user needs and supply-side capabilities in a more efficient way:
- For users, it’s time filler – filling fragmented time with precise recommendations;
- for creators, it is an equalizer of opportunity – using algorithms to make ordinary people’s content visible;
- For merchants, it’s a growth accelerator – stimulating potential consumer demand with content.
From traffic black holes to super apps, Douyin has redefined the boundaries of Internet products. Its story continues, but one thing is clear: in an era of scarce attention, whoever understands users better and connects more efficiently will become the infrastructure of the next Internet.
And Douyin is already on this road.
