TikTok: how an algorithm beat the follow graph
ByteDance launched Douyin in China in September 2016 and TikTok internationally in 2017. The product's structural innovation was building the recommendation algorithm (the “For You” feed) as the primary distribution mechanism rather than the social follow graph that Facebook, Instagram, Twitter, and YouTube had built. TikTok could show content from creators a user had never heard of based purely on engagement signals. The format produced explosive growth and content discovery patterns no incumbent had matched. By 2024, TikTok had ~1 billion monthly active users globally. The case is studied as the defining algorithm-over-follow-graph product-strategy shift.
- Story: ByteDance launched Douyin in China in September 2016 and TikTok internationally in 2017. The product's structural innovation: algorithm-driven “For You” feed instead of follow-graph feed. ~1B MAUs by 2024. US faces ongoing regulatory uncertainty.
- Why it matters: TikTok is the defining algorithm-over-follow-graph case. The strategic insight: algorithm distribution is exploratory and produces faster growth than conservative follow-graph distribution.
- Takeaway: Algorithm-driven distribution beats follow-graph for content platforms where discovery matters more than maintenance.
- Takeaway: Algorithm quality is a meaningful competitive moat — copy the format without comparable algorithm and the results are much weaker.
- Takeaway: Geopolitical risk can be a structural threat to platforms regardless of product quality.
TikTok — the four-step story
TikTok at a glance
Quick facts
The product strategy
Most social media platforms built around 2010 used the “follow graph” as the primary distribution mechanism: you follow people, you see their content. Facebook, Instagram, Twitter, YouTube, Snapchat all worked this way. The follow graph produced predictable content distribution — if you didn't have followers, your content didn't reach anyone.
ByteDance built Douyin (in China) and then TikTok (internationally) around a different primary mechanism: the recommendation algorithm. Users opening TikTok don't see content from people they follow first — they see the “For You” feed, a sequence of videos the algorithm has selected based on engagement signals (watch time, completion rate, replays, likes, comments, shares). Content from creators the user has never heard of can be shown if the algorithm predicts the user will engage with it.
The algorithm-over-follow-graph approach produced several structural advantages:
- Easier creator discovery. New creators with quality content could reach large audiences immediately. Follow-graph platforms required years of building followers before reaching meaningful audiences.
- Higher per-user time spent. The algorithm could continuously surface content optimized for individual engagement, producing higher session lengths than follow-graph feeds.
- Content variety. Users saw content from outside their follow graph, which broadened exposure to topics and creators they wouldn't have encountered otherwise.
- Lower-friction content creation. Creators didn't need existing audiences to break through. The reduced friction increased creation volume.
Format and growth
TikTok's vertical-video, short-format (15 seconds to several minutes) structure combined with the algorithm-driven distribution produced viral content patterns no incumbent had seen. Songs went from unknown to top of the charts based on TikTok adoption. Niche-subculture content reached mainstream audiences through algorithm-driven discovery. Creators built audiences in months rather than years.
TikTok acquired Musical.ly in August 2018 for ~$800M-$1B (Musical.ly's user base was largely younger US users), integrated it with the existing TikTok product, and accelerated US growth. The COVID pandemic in 2020-2021 produced massive user-base expansion as people sought entertainment during lockdowns.
By 2024, TikTok had approximately 1 billion monthly active users globally. Adjacent products (TikTok Shop for integrated e-commerce, TikTok creator funds) extended the platform into broader creator-economy and commerce categories. The format influenced incumbent platforms — Instagram launched Reels (2020), YouTube launched Shorts (2021), Snapchat launched Spotlight (2020), all explicitly modeled on TikTok's algorithm-driven short-video format.
The regulatory question
TikTok's parent company ByteDance is Chinese-owned. US regulatory and political pressure has been ongoing since 2019-2020 over data-privacy and national-security concerns. Multiple US administrations have considered actions against TikTok including divestiture requirements and outright bans. The Protecting Americans from Foreign Adversary Controlled Applications Act (signed 2024) required ByteDance to divest US TikTok operations or face a US ban.
The legal and operational status of TikTok in the US has been changing through 2024-2025 across the administration transition. The eventual outcome (ban, divestiture, continued operation) was not fully resolved as of the project knowledge cutoff. The case-study lessons about algorithm-over-follow-graph product strategy remain valuable regardless of the geopolitical outcome.
How RGM thinks about algorithm-driven product design
When clients ask about product distribution mechanisms, the TikTok case is the defining algorithm-over-follow-graph example. The strategic insight: follow graphs are conservative distribution — content only reaches people who already opted in. Algorithm distribution is exploratory — content can reach anyone the algorithm predicts will engage. For content platforms where discovery matters more than maintenance, algorithm distribution produces faster growth and higher engagement.
The harder lesson is that algorithm-driven distribution requires actually-good recommendation algorithms. TikTok's distribution worked partly because the underlying algorithm was unusually effective at predicting engagement. Platforms that copied the format without comparable algorithm quality (early Instagram Reels especially) produced significantly weaker results. We tell clients that the product-strategy choice (algorithm vs follow graph) and the engineering execution (algorithm quality) are both required — one without the other doesn't produce comparable outcomes.
Frequently asked questions
What is the “For You” feed?
TikTok's primary feed, which serves users a sequence of videos the algorithm has selected based on individual engagement signals. It's distinct from the “Following” feed (which shows content from creators the user has chosen to follow) and is the default feed users see when opening the app.
Why is TikTok's algorithm considered so good?
Several factors. The interaction signals (full-watch, replay, share, like, comment) are richer than other platforms collect at scale. The vertical-video format gives the algorithm signal density per session. ByteDance's investment in recommendation-systems research dates back to predecessor products. The combination has produced an algorithm that converts users to engaged-watching faster than incumbent platforms.
Will TikTok actually be banned in the US?
Status unclear as of the project knowledge cutoff. The Protecting Americans from Foreign Adversary Controlled Applications Act passed in 2024 required ByteDance to divest US TikTok operations or face a ban. The Biden administration enforced the ban in January 2025; the Trump administration that took office shortly after delayed enforcement and explored various deal structures. The eventual outcome (continued operation under a new structure, divestiture to non-Chinese owner, or outright ban) hasn't been finalized.
Sources & references
- ByteDance / TikTok — Product reference.
- Protecting Americans from Foreign Adversary Controlled Applications Act — US legislative reference.
- Zhang Yiming / ByteDance background — Corporate-history reference.