You do not get three seconds. You get one. Somebody's thumb is already moving before your video has finished loading, and the decision to stay or scroll happens well before they have processed what your brand is or what you sell.
That sounds bleak. It is actually quite freeing, because it means you can stop worrying about the last twenty seconds of a video until you have solved the first one.
lead with a face
The single most reliable way to slow a thumb down is a human face looking directly at the camera. It is not a trick, it is wiring: people notice people. A product on a white background has to work hard to earn the same half-second.
This is the whole reason doubleseven exists in the shape it does. When a real person fronts your content, you are not fighting for attention from a standing start every single time.
say the thing immediately
The most common mistake we see is a slow build. A logo sting, a moody establishing shot, then finally the point. Cut all of it. Whatever the payoff is, move it to the front and use the rest of the video to justify it.
If your video works with the sound off and the first frame removed, it is not a hook. It is an intro.
pace beats polish
Content that cuts every two or three seconds holds attention better than content that is beautifully lit and static. Given the choice between another hour of colour grading and three more cutaways, take the cutaways.
test the hook, not the video
When something underperforms, most brands rewrite the whole piece. Usually the body is fine and the opening was wrong. Shoot three or four different openings for the same content and run them separately. It is cheap, fast, and it tells you more than any analytics dashboard will.
Do that for a couple of months and you stop guessing what your audience responds to. You know.
