TikTok Moved the Finish Line to Seventy Percent

Listen to this episode of the Social Trend Report, then use the piece below.

Cut a video the way that won you views in June, hard hook at zero seconds, a cut every beat, out the door at fourteen seconds, and it caps out around three thousand views now. TikTok moved the number it grades you on.

The completion floor for wide distribution sat close to fifty percent for most of the last two years. It's closer to seventy now. Watch time and completion together drive something like forty to fifty percent of the ranking weight, so a video that used to clear the bar on a fast cold open now has to hold a stranger for the whole run, not just the first two seconds.

Oracle is grading the homework now

The Oracle-led joint venture that took over TikTok's American operations finished the handoff early this year and has spent the months since retraining the recommendation engine on nothing but US-sourced signals: watch time, shares, saves, search queries. The retrain runs through the middle of 2026 and touches something like a hundred seventy million American accounts. Your reach moved because the model grading you got new homework, and it's still learning it.

The mechanics, per SocialPilot's breakdown of the 2026 ranking signals and echoed across independent creator-economy trackers running their own test accounts: completion under seventy percent caps a video inside its first test batch. Clear it, and the algorithm keeps widening the audience. A sixty-second video used to earn wider distribution at roughly thirty seconds of watch time. Now it needs something closer to forty-two.

The same number shows up across separate trackers running separate accounts, which is about as close to confirmed as anything gets on a platform that publishes no official change log.

What the social side does this week

Pull your last twenty posts. Not the ones you're proud of, all twenty. Sort by completion rate, not views, not likes. The pattern shows up before you finish the spreadsheet: the videos built to front-load a hook and bail are the ones losing reach, including some of what used to be your best performers.

Length isn't the fix. Shape is. Build a video that needs its ending, not a headline that already gave the point away in the first two seconds. Move your best line to the last three seconds instead of the first three. Make somebody sit through the setup to earn it.

Pasteable copy

This one's a post, not a reply. Practitioners talk about platform shifts like this openly, and being early and honest about one is content on its own:

"We re-cut our last month of video against completion rate instead of views. What we thought was working wasn't. Full-watch videos and quick-hit videos got about the same reach a year ago. Not anymore. Blame the new floor, not your team. Happy to compare notes."

Still direction

Shoot the editing bay, not the feed. A phone propped against a stack of raw footage on a desk, the timeline scrubber mid-drag on the laptop screen behind it, natural window light, nothing on the monitor sharp enough to read. One frame, handheld feel, nothing staged. The still says you're actually re-cutting, not just posting about it.

Fifteen to thirty second video, shot today

Vertical, one operator, phone propped against a laptop. Five shots:

1. Two seconds. Hand scrubs a laptop timeline, close, no readable text.
2. Five seconds. Person on camera, mid-shot: "Our reach dropped and we finally found out why."
3. Six seconds. Close on a notes app or whiteboard: fifty percent crossed out, seventy percent written under it. Nothing branded.
4. Six seconds. Back to the person: what changed in the edit, in one sentence. The best line moved to the end instead of the front.
5. Four seconds. New cut of the video, playing, full screen. Hold on it. No card, no logo sting, no music swell.

Post it once, in feed, no boost. The credibility is in the fact that it looks like it took the twenty minutes it took.

Where this goes

The retrain doesn't finish until the middle of 2026, so the floor probably isn't done moving. Bank on the completion math getting stricter, not friendlier. Full-watch signals are exactly what an advertiser-funded platform wants more of. The accounts that get ahead of it are the ones grading their own videos against that yardstick before they hit publish, not after the numbers come back thin.

This is the extension of this week's episode. The show is the briefing, the article is the detail, the post is what you actually ship. Subscribe to the Social Trend Report on Apple Podcasts or Spotify and it's in your ears before the next re-cut.

If you want a second set of eyes on your own completion numbers, find me on LinkedIn.