For You Page: how TikTok decides

The FYP is a ranker, not a magic feed.

TikTok's For You Page is a probabilistic recommendation system that ranks every post against a candidate viewer based on a weighted combination of signals: completion rate, watch time, shares, comment-and-reply patterns, save rate, and prior viewer-creator affinity. None of this is hidden machinery; TikTok itself describes the high-level architecture in its Creator Academy. What matters operationally is the order of weights — completion rate sits at the top, swipe-away rate at the bottom, and everything else fights for the middle. Treat the FYP as a graded test on each post, not a lottery.

The 200-viewer seed pool — your first 30 minutes.

Every new post is shown to a small cold pool first, typically around 200 viewers selected to match the inferred niche of the account. If the seed pool's completion rate, share rate, and comment-and-reply rate beat the rolling threshold for your account's category, the video gets promoted to a second-tier pool (1,000-5,000 viewers). The first 30 minutes after publish are the cold-start window: this is when the seed pool is being scored. A video that does not clear the seed-pool threshold rarely recovers — pushing it later (via Promote or sharing) helps less than people assume.

Top positive signals — completion is king.

Completion rate is the single most heavily weighted signal: a 15-second clip watched to the end is worth far more than a 60-second clip watched to 40%. Below completion, in decreasing weight: shares to other apps (a strong intent signal), comments that receive replies (interaction depth, not just count), watch-time anomaly (rewatches, the viewer pulled the timeline back), and save rate. Saves matter most for utility content (tutorials, recipes) and less for entertainment. Engineer your hook to land before the 2-second mark and your payoff to land before the swipe-away decision.

Negative signals — the buried-video patterns.

Three negative signals actively suppress distribution: rapid swipe-aways inside the first 2 seconds (the hook failed), 'Not interested' taps (an explicit downvote), and reports. Of the three, swipe-aways are by far the most common and the most fatal — they tell the algorithm the video is not matching its candidate viewer. The fix is not louder hooks; the fix is tighter audience targeting through your niche signal. A video that swipes-away 70% in the seed pool was shown to the wrong people, which means your account's niche signal is muddled.

Why a video doesn't get pushed — the shadowban myth.

When a creator says 'my video got shadowbanned,' they are wrong about 95% of the time. The actual story, almost always: the seed pool gave the video a poor completion rate, so it did not get promoted to a wider pool. This is normal algorithmic behavior — not punishment, not a flag, not a manual review. Real shadowbans (where a creator's content is intentionally limited at the account level) do happen, but only after a Community Guidelines violation, and they show up clearly in the analytics tab. Confusing a low-performing post with a shadowban wastes weeks chasing a phantom problem.

Diagnosing why a video flopped — a numbered investigation.

Run this exact sequence the next time a video underperforms: (1) Check completion rate first — anything under 30% on a sub-30-second video means the hook failed. (2) Check the swipe-away curve — if more than 50% of viewers exited in the first 3 seconds, the audience match was wrong; if exit was steady through the middle, the payoff was weak. (3) Check shares and saves — if the engagement-to-view ratio is below 1%, the video did not earn outward signal. (4) Check whether this video matches the niche you have been training the algorithm on for the past 10-20 posts — if it does not, niche drift is the cause. (5) Only after all four checks should you consider whether Community Guidelines triggered a real distribution limit; the analytics tab will say so explicitly.

Niche signal and the slow drift.

TikTok's classifier infers your account's niche from the first 10-20 posts that pass the seed-pool threshold; once locked, the niche signal decides who composes the seed pool for every future video. Drift happens when a creator chases trends across categories — gaming Monday, dance Tuesday, recipes Wednesday — and the visible symptom is a flat-then-declining view curve across 4-6 weeks with no single bad post explaining it. The algorithm is reclassifying you into a weaker bucket and rebuilding the seed pool from scratch. Recovery takes 2-4 weeks of disciplined on-niche posting. The agency rule: one cross-niche experiment per month, fixed schedule, signposted in the calendar.

Reset signals — delete, warm-up, pin discipline.

Delete a low-performer only within the first 24 hours after publish, and only if it was clearly off-niche or contained an error (wrong audio, broken caption, off-brand framing). Deleting after 24 hours can actually hurt — the algorithm reads it as a signal of niche-commitment failure. For a stale profile (no consistent posting in 30+ days), the warm-up is a 7-14 day stretch of daily on-niche posts to re-establish the classifier read. Pinned posts are also a niche signal; if the top 3 pinned posts are off-niche, unpinning them helps the warm-up land faster.

LIVE notification push — who actually gets the alert.

TikTok does not push your LIVE notification to all your followers. The push goes to a subset selected by recent engagement: followers who have watched one of your videos to completion within the last 7 days, plus a smaller cohort of frequent LIVE viewers regardless of follow status. This is why a creator with 100K followers may only see 800-1,500 LIVE-push entrants in the first minute. The lever you actually control: short-form engagement in the 7 days before each LIVE. A creator who keeps a healthy short-form drumbeat earns a larger LIVE push than a creator who only streams.

The agency's weekly profile-health review — a 5-metric scan.

Every Monday, before the calendar meeting, the agency runs this 5-metric scan on each managed creator: (1) Average completion rate across the last 10 short-form posts — flag anything below 35% on a sub-30-second account. (2) Posting cadence — under 4 posts per week is a warm-down signal; the algorithm reads it as niche disengagement. (3) Niche drift score — proportion of last 10 posts that match the primary niche; below 70% triggers a niche-correction plan. (4) Engagement-to-view ratio — combined likes/comments/shares per view; below 4% means the audience match is muddied. (5) LIVE push rate — fraction of followers who got the LIVE-start notification on the last 5 streams; below 1% means short-form engagement is too thin to feed the push.

Common myths debunked — and the operational takeaway.

Three myths the agency rebuts every week: 'deleting flops cleans my account' (false after 24h — reads as niche-commitment failure); 'posting time-of-day is decisive' (false — third-order vs completion rate; spending energy on the hook is 10x more productive than spending it on the clock); 'hashtag stuffing helps' (false — 3-5 relevant niche tags is enough, more signals desperation). The operational takeaway: you cannot manage 'the algorithm,' but you can manage the seed pool post by post — hook in the first 2 seconds, payoff before the swipe-away decision, niche consistency across the last 10 posts, completion rate as the north-star metric. Do that for 30 days and the FYP stops looking like a black box.

Black Ads Agency runs feeder-short experiments aligned to this FYP weighting on all 5 partner servers — MENA (Faction 108135), FR+ (Faction 115414), IT (Faction 117633), DE (Faction 120935), US (Faction 128508). BlackOS logs completion rate, watch time, and Diamond-velocity carry from each post into the next TikTok LIVE, with a 0% commission ledger that keeps creator-side economics clean.