What TikTok's algorithm actually measures
TikTok has never published its complete algorithm. But through analysis of millions of videos, researchers and creators have identified the most important signals. Here are the 6 that carry the most weight.
Signal 1: Completion rate (the most important signal)
What percentage of your video do people watch on average? A video that's watched 100% or even multiple times gets more reach automatically. TikTok calls this internally "re-watch". Goal: above 70% completion rate.
Tactic: Keep videos short (15–30 seconds), end with a cliffhanger for the next episode.
Signal 2: Shares (strongest engagement type)
Being shared is the strongest signal to TikTok: someone finds your video so valuable they want to show it to someone. Shares count 5–10× more than likes. Create content that people would send to their friends.
→ Factors that generate shares
Signal 3: Saves
Saved videos signal: "I want to watch this again later." Especially important for how-to content, tutorials, and list-style videos. Create content so useful that people save it.
Signal 4: Comments (quality > quantity)
A comment like "I tried this and it worked!" signals more than 10 "😂" comments. Real reactions showing the video was useful or touching are gold.
→ CTAs that generate real comments
Signal 5: Profile visits after the video
When someone visits your profile after your video and watches more, it's a strong signal: you've gained a new fan. Always build a reason to click "more" at the end of your videos.
Signal 6: Posting time and frequency
The algorithm favors channels that appear regularly. Channels posting daily at the same time get more preview impressions. Irregular channels drop out of the promotion rotation.
→ Calculate best posting time | Create posting schedule
What the algorithm penalizes
- Low completion rate: Under 40% — barely distributed further
- Imported videos: Watermarks from other platforms → reach penalty
- Too many hashtags: More than 5–7 hashtags reads as spam
- Inactive periods: 2+ weeks without posting → algorithm "forgets" you
The strongest algorithm hack: daily consistency via AI
TikTok's algorithm loves consistent creators. ClipDraft generates new, unique videos daily — without you having to invest daily time. This keeps you permanently in the algorithm's rotation.
FYP Mechanics: How the For-You Page System Makes Decisions
The For-You Page works as a multi-stage funnel. Every video goes through several distribution rounds:
- Round 1 (~200–500 accounts): Small test group from your niche. If engagement is good: advance to Round 2
- Round 2 (~2,000–5,000 accounts): Broader niche audience. Good performance advances further
- Round 3 (~20,000+ accounts): First major reach — here it's decided whether a video trends toward viral
- Round 4+ (unlimited): For videos still performing strongly in Round 3 — no fixed ceiling
This explains a fascinating phenomenon: TikTok videos can suddenly go viral months after upload, when the algorithm shows the video to a new target audience.
Algorithm Updates 2026: What Has Changed
TikTok has significantly updated its algorithm in the last 12 months:
- Quality score introduced: TikTok now actively evaluates content quality — videos with "low information value" receive less reach
- Diversity factor increased: Creators see more variety in their FYP — the algorithm shows fewer repetitive content types from the same creator
- Search signal weighted more heavily: Keywords in captions and spoken text are more heavily counted for search algorithm distribution
Video Completion Data: Where Viewers Actually Drop Off
Based on creator analytics data, clear patterns emerge where viewers abandon videos:
- Seconds 0–3: Weak hook → 40–60% of viewers scroll away
- Seconds 10–15: Second critical point — without a new information element, attention drops
- Last 20% of video: Drop-off rate is lowest here — viewers who watch until near the end usually watch through
Consequence for video structure: hook in seconds 1–3 → new information impulse every 10–15 seconds → strong ending (conclusion, CTA, or cliffhanger).
Testing Methodology: How Creators Systematically Learn the Algorithm
The most effective way to understand the algorithm: systematic A/B testing.
- Isolate one factor: Test only the hook — same content body, three different openings as separate videos
- Evaluate after 48 hours: Which hook had the highest 3-second rate? That's your winning pattern
- Replicate the pattern: Use the winning hook type for the next 10 videos
- Test the next factor: Video length, sound type, caption format, posting time
After 60 videos with systematic testing, you have a creator-specific algorithm profile: you know what works for your specific audience — and that's more valuable than any general recommendation.
Frequently Asked Questions
Why do some videos perform well weeks after upload?
TikTok continuously re-evaluates videos. As your account profile changes (more followers, better engagement rate), the algorithm can show older videos to a new, better-matching test group. External traffic spikes (someone shares the video on Twitter) can also reactivate the algorithm. Conclusion: never delete videos — they can be catapulted through the funnel at any time.
Is the TikTok algorithm different for new channels vs. established ones?
Yes — but not necessarily worse for new channels. New accounts often receive more test reach because TikTok needs to classify them. This explains the "honeymoon effect": the first 5–10 videos of a new channel often get disproportionately high reach. The difference is that established channels have a follower base that's contacted first in the test phase — which increases the engagement rate.
Can I manipulate the algorithm with follow/unfollow or engagement pods?
Short-term maybe — long-term no. TikTok recognizes unnatural engagement patterns (e.g., 500 likes in 5 minutes from inactive accounts) and penalizes them with reach reduction or account restrictions. Follow/unfollow also explicitly violates TikTok's Community Guidelines. Sustainable growth only comes from genuine, organic engagement.