short-form-what-works

name

short-form-what-works

description

Evidence-based rules for what actually moves short-form video performance (TikTok, Reels, Shorts), distilled from Layers Research's 2026 TikTok Posting Report (23M videos, 13.3T views). Use this skill whenever the user asks about posting strategy mechanics — best time to post, posting frequency, how many posts per day, spacing, taking breaks, video length, slideshows vs. videos, caption length, hashtags, resolution/aspect ratio, deleting flops, shadowbans, follower counts vs. reach, or what to post after a viral hit or a flop. Also trigger when the user repeats posting folklore ("post at 6pm", "don't post twice a day", "keep it under 30 seconds", "the algorithm punishes X") so their plan can be checked against the data, or when auditing a posting schedule or content calendar.

Short-Form: What Actually Works

Data-backed posting rules from Layers Research's 2026 TikTok Posting Report — 21 pieces of industry folklore tested against 23 million TikTok videos and 13.3 trillion views. Every finding is a within-creator comparison (each video ranked only against the same account's other posts), so audience size and niche are controlled away. Use this to overrule folklore with effect sizes. Pairs with the shortform-content skill (Kallaway's Lego Brick framework): that skill covers how to make the video; this one covers everything around it.

Units: effects are in "spots" — percentile points within a creator's own catalog. Near the middle, +4 spots ≈ +14% views, +10 ≈ +42%, +20 ≈ double. Payouts are jackpot-shaped: a 90th-percentile post earns ~8× your median; a 99th, ~60×.

The headline

When you post is noise. What you post is signal. Timing plus account features predicts a creator's hits at AUC 0.532 (barely above coin-flip); reading the content lifts it to 0.574. All 168 hour-of-week slots combined explain 0.007% of performance. Redirect every minute of calendar optimization into the content itself — especially the first two seconds.

The levers, ranked by effect size

Actual levers, largest first (view-equivalents for a typical mid-catalog post):