The first 3 seconds decide if your ad gets watched

The ad is usually good. That's the part that makes this conversation awkward. Someone paid for a shoot, the offer is genuinely strong, the edit is clean, the captions are burned in - and the reporting says almost nobody stayed long enough to hear any of it.

Impressions in the tens of thousands. Average watch time under two seconds. Spend that looks like it bought reach and bought nothing.

The instinct at that point is to rewrite the script, change the offer, or blame the targeting. Sometimes that's right. Far more often, the ad never got a chance to be judged at all. The viewer made a decision before the pitch started, and everything downstream - the demo, the proof, the call to action - was written for an audience that had already gone.

What follows is what actually happens in those opening seconds, what the platforms are measuring while it happens, and what you can change without booking another shoot.

---

1Three seconds isn't a figure of speech. It's a line in your reporting.

"The first three seconds matter" gets repeated so often that it sounds like advice from a poster. It isn't. Every major platform has a measurable threshold in the opening seconds of a video, and the numbers on either side of that threshold are what your account is being optimised against.

They're worth knowing exactly, because they're all different:

Two things fall out of that table.

The first is that your opening seconds are not just creative preference - they're the input to the algorithm's opinion of your ad. On Meta, ThruPlay counts a play to completion for videos under fifteen seconds, and at least fifteen seconds for longer ones, which means everything you're paying for is gated behind an opening the viewer chose not to leave. On TikTok, focused view optimisation deliberately looks past the cheap two-second impression and chases people likely to watch six seconds or interact within them.

The second is more uncomfortable. On YouTube's skippable format the viewer is handed an exit at five seconds, and the platform doesn't charge you on CPV bidding unless they stay to thirty or interact. A skipped ad is close to free. It is also close to worthless, and a campaign full of them looks perfectly healthy on spend while doing nothing at all.

---

2You're not competing with other ads. You're competing with the swipe.

Most ad briefs are written as though the viewer is comparing options. They aren't. Nobody in a feed is weighing your roofing ad against another roofing ad. The only competition is between watching anything at all and moving the thumb.

That changes what the opening is for. It isn't an introduction to your business. It's a bid for permission to keep going, made against a default of "no", and it has to be won before the viewer has consciously decided anything.

The practical consequence is that the opening has to work at the level of a thumbnail, not a paragraph. In the fraction of a second before anything else registers, the viewer takes in movement, a face or not a face, and whether there's text they can read at a glance. Whatever your first frame contains is what they're judging - which is why a black frame with a logo fading up is one of the most expensive edits in advertising.

---

3Most hooks are an introduction wearing a hook's clothes

Here's the pattern we see most: the first three seconds contain a logo sting, a drone shot of the building, or a person saying "Hi, I'm Dave from Dave's Plumbing, and today I want to talk to you about…"

None of that is bad footage. All of it is runway. It's the part of the video where the ad is getting ready to say something, and the platform is billing you for it while the viewer leaves.

The fix is almost mechanical. Find the sentence in your script that a stranger would actually stop for - usually the third or fourth one, the one where you finally name the problem or the number - and start there. Everything before it goes, or moves to the middle where it costs you nothing.

Some openings that consistently earn the next three seconds:

• The specific claim. "We replaced this roof in one day" beats "quality roofing you can trust", because the first is checkable and the second is wallpaper.
• The visible problem. Open on the mess, the leak, the before. People stop for a problem they recognise well before they stop for a solution they don't.
• The named audience. "If you own a rental in Houston" filters ruthlessly and rewards you with the right viewers - and a smaller, better-qualified hook rate is often the more profitable outcome.
• The mid-action open. Start with something already happening on screen. Motion in frame one is the cheapest attention you will ever buy.

---

4Everything you wrote in the first three seconds was heard by nobody

Feeds are muted by default, and a large share of your audience will never turn the sound on. If your hook is a spoken line, the hook doesn't exist for those people. The information simply isn't delivered.

This is the single most common gap between how an ad performs in the edit suite and how it performs in the account. Reviewed on a laptop with sound, the opening lands. Served into a silent feed, the same three seconds are a person moving their mouth.

Three things that close the gap, in order of impact:

1. Put the hook on screen as text, in frame one. Not as a caption that appears when they say it - as a title card that is already there. Six words or fewer, large enough to read at arm's length on a phone.
2. Caption everything else. Burned-in subtitles, not platform auto-captions, which get toggled off and don't render consistently across placements.
3. Keep text out of the interface. Reels, Shorts, and TikTok all overlay usernames, captions, and buttons across the bottom and right of the frame. Text placed there is text the viewer never sees. Keep it in the middle band and check each placement in preview, not just the one you built for.

---

5A hook that doesn't match the ad buys you the wrong three seconds

Once people learn that hooks are the lever, the next failure arrives quickly: a hook engineered purely for stops. A shocking statistic with nothing to do with the service. A "wait for it" that pays off with nothing. Someone dropping something.

These work, in the narrow sense that hook rate climbs. Then everything after it gets worse. Retention collapses at the moment the ad reveals what it's actually selling, click-through slides, and cost per lead rises while the top-line video metrics look better than they ever have.

You've bought attention from people who wanted the thing in the hook, not the thing in the ad. Worse, on platforms that optimise delivery around engagement signals, you've now taught the algorithm to go and find more people like them.

The test is simple and it's a writing test, not an editing one: would the hook still make sense to someone who ends up buying? If the opening line and the offer belong to two different ads, the hook is a tax, not an asset.

---

The cheapest test in advertising

The last thing worth changing is the unit you test. Most accounts test whole ads against whole ads, which is slow, expensive, and tells you almost nothing about why one won.

Test openings instead. Take the ad you already have, cut four different first-three-seconds onto the identical body, and run them against each other. Same footage, same offer, same landing page, four doors into it. The edit is an afternoon. The read is clean, because only one variable moved.

A few things that keep this honest:

• Judge on hook rate first, hold rate second, cost per result last - in that order, because they diagnose different failures.
• Give each variant enough impressions to mean something before you call it. A day of data on a small budget is noise.
• Keep the winner and change one thing again. Hooks decay as an audience sees them repeatedly; a library of proven openings is worth more than any single ad.

---

The version you can do this week

If you do nothing else from this article, do these five things before your next campaign goes live:

None of that requires new footage, a bigger budget, or a different agency. Most of it is an hour in the timeline and ten minutes in the ads manager.

That's the pattern worth keeping. The expensive part of a video ad - the shoot, the offer, the proof, the call to action - is almost always fine. It's just sitting behind three seconds nobody agreed to watch.

All articles