The Window Is The First Five Days

The view stopped being the unit and most people are still optimising the metric that no longer decides anything. Behaviour inside the view is weighed at 4 checkpoints, tracked independently: 1 s, 3 s, the mid dip and the end. 1 s is settled by the visual, before the audio has a chance to matter. 3 s is the commit, and it is where a murky hook loses the person it just stopped. The mid dip is where the delivery either matches the promise the hook made or quietly breaks it. The end is finish, rewatch or share, and it is the only 1 of the 4 that compounds. 4 readings, 4 different ways to die, and 1 average that hides every 1 of them. 2 clips can both sit at 62 per cent and be nothing alike: 1 holding 95 at the first check and falling to 20 by the last, the other flat at 60 the whole way through. The first 1 gets shared 1 more time per 100 viewers and the second gets nothing, and no dashboard that reports an average will ever tell you which 1 you made. Read the 4 curves, not the 1 number.

A hook now has 2 jobs and the second 1 is the 1 nobody was taught: stop the scroll, then signal the niche so the routing goes to the right viewers. A hook that stops a stranger but confuses them collects views from the wrong audience, and the wrong audience leaves at the 3 s mark because the promise was made to somebody else. 63 per cent of top videos hook inside 3 s (TikTok for Business), and the average mobile decision takes about 1.7 s, which means the visual has to be legible before a single word of the script is spoken. That is why a hook that takes 5 s to become readable is not a slow hook, it is a hook that never landed at all. 2 jobs, 1 artifact. The first frame names the category, the first line names the audience, and the promise has to be specific enough that the right person recognises themselves inside 1.7 s. Warmth does this better than volume, because this audience reads meaning faster than it reads polish.

The thresholds are the useful part because they are pass or fail rather than directional. 70 per cent plus intro retention is strong. Under 40 per cent is dead on arrival, and no amount of mid-video craft rescues it. 63 per cent of the top performing videos hook inside 3 s. Vertical 9:16 now accounts for 60 per cent plus of all social video, reels carry 50 per cent of the time spent on Instagram across 2 billion monthly users, and a reel reaches roughly 2 times the audience of a static photo posted to the same account. Length behaves the way most people guess wrong: a 12 second cut outperforms a 60 second cut, because a short film watched twice beats a long 1 watched once, and 7 to 15 seconds is where that arithmetic lives. So the shape of a working cut is knowable before anything is shot. 9:16, 12 seconds, hooked inside 3, with a mid point that delivers what the first second promised. 4 numbers, 2 thresholds, 1 length band.

Velocity speed ramps are still the 1 technique nobody has replaced, and the reason is the ease rather than the speed: an S-curve that snaps in and eases out reads as camera work, while a linear ramp reads as a slider being dragged by hand. Cut frequency sits at 0.5 to 2 seconds in vertical, which is faster than horizontal because the frame is taller and the eye travels further between cuts. Captions are mandatory rather than helpful, set at 36 to 60 point, keyword highlighted, centre or upper third, because most of the audience watches with the sound off 1 minute and on the next, and the caption is the second audio track. Cinema grading stopped being a separate lane and moved into every platform: short-form content is graded now, not just cut, and sound design plus motion graphics are baseline rather than garnish. 1 rule sits under all 4 of those. The technique is invisible when it is right. The frame should read as composed rather than treated.

The view stopped being the unit. Behaviour inside the view is the unit.

There are 2 clocks running under every trend and they run at different speeds. Audio trends peak in 7 to 10 days and then the distribution shifts underneath them. Format trends, the split screen, the text overlay, the carousel, run 3 to 4 weeks. The useful window is the overlap, and it is the first 3 to 5 days, which is short enough that the decision to jump is the entire edge. The fading signals are visible early and cheap to read: brands arriving in numbers, and the same format showing up from 20 plus accounts. When both are true you are not early, you are the audience. The arbitrage here is arithmetic rather than taste. A 40 per cent head start costs nothing but 3 days of not deciding, and it is the difference between being routed as the original and being routed as the 21st copy of somebody else's format.

The last trend is the 1 that is not about the platform at all. Meaningful beats addictive, warmth and specificity beat hype, and that reading was in the audience research before the algorithm changed to reward it. The single most watched piece of this period is a 336 million view single-take lip sync filmed by a truck driver with no production value in it, and that is not an argument against craft, it is an argument against polish used as a substitute for 1 specific promise. The 2016 nostalgia wave is trending for the same reason: people want the energy of cracking a system open rather than the energy of performing inside it. The probability that a specific promise finds the right 1 person is not zero, and most people are pricing it at zero, which is why so much work dies at 3 s holding a beautiful frame with nothing on the line. Hook in 1 s, cut on the beat, caption everything, and lead with warmth plus 1 specific promise. 4 moves, no hype, and the same argument at length runs 393 pages in the book published in April 2026.