TikTok's recommendation system is the most powerful content distribution engine ever built. We break down every ranking factor, from completion rate to sound interaction, so you can engineer viral reach systematically.
Every video goes through a three-stage distribution process. Understanding each stage is key to engineering reach.
Every new video is shown to a test pool of approximately 200 diverse viewers — a mix of your followers and non-followers. The algorithm measures early signals intensely in this window. A zero-follower account gets the same initial opportunity as a million-follower creator.
Completion rate, likes, comments, shares, and rewatches are scored and weighted by a machine learning model. Completion rate holds the most weight — a video watched all the way through is a strong signal of quality. Rewatches are an even stronger signal, indicating exceptional value.
Videos that pass the initial score threshold are shown to progressively larger audiences: 5K, then 50K, then 500K, then millions. Each expansion is another test — the algorithm checks if signals hold up at scale. Videos that fail at any threshold stop expanding.
These signals are scored and combined by TikTok's recommendation model. Weights are based on documented creator experiments and platform disclosures.
Completion rate directly signals that your content delivered what the hook promised. TikTok's model is trained on the simple idea that if a viewer chose to watch until the end — and especially if they rewatched — the content was genuinely valuable to them. No other signal carries this same level of unambiguous user intent. A like can be habitual. A rewatch cannot.
Follow the exact path your content takes from upload to potential virality — or suppression.
Video is processed, transcribed, and categorized by topic, audio, and visual content. Metadata is extracted.
Automated checks for spam signals, community guideline violations, watermarks, and minimum quality thresholds. Fails here = never shown.
Shown to a diverse sample: some followers, some interest-matched non-followers. Algorithm watches every interaction closely in this window.
Completion rate, engagement rate, share rate, and rewatch rate are measured and scored against category benchmarks.
Composite score is calculated. If above threshold, expanded distribution is triggered. If below, video enters a holding pool and may resurface later.
5K → 50K → 500K → Millions. Each tier is re-evaluated. Strong signals = continued expansion.
Distribution stops. Video remains visible on profile but receives no algorithmic push. May resurface after 90+ days.
| Factor | Weight | Optimization Tip |
|---|---|---|
| Completion Rate | ⭐⭐⭐⭐⭐ | Hook viewers in the first 0.5 seconds — use a surprising visual, bold text, or an unanswered question |
| Rewatch Rate | ⭐⭐⭐⭐ | Create loop-worthy endings that connect back to the opening frame — viewers rewatch without realizing it |
| Shares | ⭐⭐⭐⭐ | Create content people want to send to specific friends — "tagging" content outperforms broadcast-style content |
| Comments | ⭐⭐⭐ | Ask open-ended questions or make safe but provocative statements that invite debate |
| Likes | ⭐⭐⭐ | Prompt engagement at peak emotional moments — when the payoff lands, add a verbal or visual CTA |
| Profile Visits | ⭐⭐ | Maintain a strong, consistent creator identity that makes viewers curious about your other content |
| Video Downloads | ⭐⭐ | Create reference content — recipes, tutorials, data summaries — that people want to save for later |
"Hashtags determine who sees your video — more hashtags mean more reach."
Hashtags help with topic categorization but completion rate is 10x more impactful on distribution. Over-hashtagging can actually signal spam-like behavior.
"You need 10K followers before TikTok will push your content to non-followers."
TikTok regularly pushes accounts with zero followers to hundreds of thousands of viewers if their signals are strong. Follower count has minimal weight in FYP distribution.
"Posting at the exact right time is the most important factor for reach."
Signal quality in the first 30 minutes after posting matters far more than the exact posting time. A video with weak signals at peak hours will still underperform.
How each ranking signal contributes to overall algorithmic distribution score, based on documented creator experiments.
Run through this list before publishing every video.
Use our testing framework to systematically experiment with your hooks, pacing, and CTAs to find what drives completion rate on your account.
View Testing Strategy