Grow your TikTok with real accounts.

TikTok’s For You Page is an early-window-heavy ranker. The first 2–4 hours after upload decide whether a video sees 1,000 views or 1,000,000, and the signals the ranker weighs during that window are substantially different from the ones it tracks later. The clearest public confirmation of this came in the 2021 New York Times disclosure of TikTok’s internal algorithm document — the leak walked through how the ranker scores per-video signals over fixed early-impression windows. TikTok itself has been unusually open about the system shape on its own Newsroom “How TikTok recommends content” explainer. Understanding the window shape is how creators either break through or watch video after video flatline at low view counts.

Watch-through percentage in the first 200–400 impressions is the decisive early signal. If 70%+ of that probe audience finishes the video, the FYP ranker reads it as high-retention content and expands distribution in steep lifts — typically 10x, then 10x again, in roughly 45-minute increments. Falling below 50% watch-through in that window effectively caps the video’s ceiling no matter what happens later. The 1-second view threshold matters here too: TikTok counts a view at 1 second of watch time, but the ranker reads completion rate (full play-through) as a separate, much stronger signal. A video with 10K 1-second views and 200 completions is a flop; 2K views with 1,500 completions is a hit waiting to expand.

Re-engagement acts as a multiplier on top of watch-through — rewatches, plus the specific shape of what follows. Saves weight heavier than likes; shares weight heavier than saves; comments from accounts the viewer has never interacted with weight heaviest of all, because the ranker reads “new connection formed” as a higher-value signal than repeat engagement. Like-share-comment ratios that veer too far from baseline (a video with 50,000 likes but 3 comments and zero shares) trigger the ranker’s low-quality-engagement heuristic and distribution dampens fast.

Regional bias matters more on TikTok than on Instagram or YouTube. The ranker reads the account’s language tag, the device region, and the IP-level geography of the probe audience, and strongly prefers to keep a video inside its originating region for the first distribution wave. A US account seeded to EU viewers during the probe window sees distribution dampen quickly — the ranker reads the low-engagement probe data as content-fit failure when it’s actually geographic-fit failure. This is why geo-verified follower composition matters for TikTok specifically, and why scattered follower mixes leave distribution on the table.

Follower count doesn’t drive first-wave distribution, but it heavily shapes the replay wave — TikTok’s loop of re-surfacing a strong video to viewers who followed the creator after the first surge. Accounts under 10,000 followers rarely see that second wave kick in meaningfully; past 50,000 it becomes a reliable tailwind. The Creator Fund threshold sits above both: 10,000 followers, 100,000 video views in the last 30 days, and 18+ account age. Creator Fund payouts are famously low (roughly 2–4 cents per 1,000 views depending on region) but the real upside is the analytics-dashboard access that comes with qualification — the only way to see exactly what the FYP ranker is reading on your videos.

Shadowbans on TikTok are real but mis-attributed. A “shadowban” in the creator vocabulary is usually a distribution dampening triggered by one of three signals: a single low-watch-through video pulling the account-level baseline down, a community-guideline strike on a recent upload (visible in the support inbox), or a sudden spike in followers/likes from an unvetted source that flagged the anti-spam heuristic. The third is what happens when creators buy bot-tier engagement. TikTok’s own Transparency Center reports publish the volume of accounts and videos taken down each quarter under coordinated-inauthentic-behavior policies — the bot-tier networks consistently dominate that data. The first two distribution-dampening modes are content-side fixes. Real-account orders paced inside the probe window don’t trigger the heuristic; the velocity curve looks like organic discovery, which is what the heuristic was tuned for.