Prepublish Research

Findings computed from public YouTube data. Every number lists its sample size, date range, and source, and both studies publish their row-level data.

Studies

The 349-video hook study

Published 14 July 2026 by Atakan Atamert

Do the first 45 seconds of a channel's biggest hits differ from those of its biggest flops?

Prepublish analyzed the first 45 seconds of 349 YouTube videos across 36 channels (177 channel overperformers at a median 4.7x normal views, 172 flops at 0.28x, 254 million combined views). Speaking pace (181 vs 180 words per minute) and time to first payoff (word 14 vs 15) showed no difference between hits and flops. Flops were 60 percent more likely to give no reason to keep watching; hits carried specific numbers (63 percent vs 52 percent) and matched their title's promise (87 percent vs 78 percent) more often.

Read the study · Download the row-level CSV

The small-channel replication

Published 24 July 2026 by Atakan Atamert

Do the hook study's patterns hold on channels with 1,000 to 20,000 subscribers?

A preregistered replication (predictions frozen at osf.io/2zq9p before data collection) tested the 349-video hook study's patterns on 96 channels with 1,000 to 20,000 subscribers, 941 videos total. None of the three patterns replicated, with 94 to 98.5 percent power to detect the original effect sizes. The one repeating signal: greeting openings leaned toward flops (23.3 percent vs 16.6 percent).