Everyone agrees the opening matters. Almost nobody agrees on what "the opening" is, how long it lasts, or what you're supposed to put in it. The result is a lot of ads that front-load motion for its own sake and bury the brand until second four.
The research is more specific than the folklore, and it points somewhere slightly unexpected.
The window is shorter than three seconds
The method is the interesting part. Every ad was tested twice: once as made, and once as a "bad twin" with the brand elements deliberately stripped out. That isolates branding from everything else — the same footage, the same pacing, the same edit, minus the distinctive assets.
The stripped versions did substantially worse. Well-branded ads were 2.5× more effective in low-attention settings, and one brand code returned 3.5× the attention-adjusted ROI of its unbranded twin. The study put the waste from poor brand integration at roughly 66p in every £1.
So the useful reframing of "the 3-second hook" is not how do I stop the scroll. It's in the 1.5 seconds I get, is the brand present enough to be encoded? A hook that wins attention and spends it on an unbranded visual has bought something it can't keep.
Value accrues before engagement does
A separate Facebook/Nielsen analysis of 173 Brand Effect studies found meaningful lift from views of under three seconds: 47% lift in ad recall, 32% in brand awareness, 44% in purchase intent, from the impression alone.
That analysis included only campaigns that showed positive results, which is a selection effect big enough to matter. It is evidence that value can accrue before engagement, not evidence of how much you should expect. Use it to argue that the opening frame is doing real work; don't use it to forecast lift.
What this means for the first frame
Three practical consequences, in order of how much they change output:
1. Brand in the opening, not the ending
The instinct to "earn" the logo by withholding it is inherited from television, where the viewer is captive for the full thirty seconds. In a feed they are not. If your distinctive assets — colour, typeface, product silhouette, a recurring character or format — don't appear inside the first couple of seconds, the majority of exposures carry no brand at all.
2. Hook rate is a diagnostic, not a goal
Hook rate (three-second views over impressions) tells you whether the opening held anyone. It doesn't tell you whether holding them was worth anything. An ad with a strong hook rate and weak hold rate has an opening that writes a cheque the rest of the ad can't cash — usually a visual gag or a shock cut with no relationship to the product.
Read the pair together:
| Pattern | What it usually means | Fix |
|---|---|---|
| Low hook, low hold | The opening isn't landing at all | Replace the first two seconds outright |
| High hook, low hold | The opening is unrelated to the payoff | Keep the open, rebuild the middle to deliver on it |
| Low hook, high hold | Good ad, bad doorway | Cheapest win available — re-cut the open only |
| High hook, high hold | Working | Vary the open, keep the structure |
3. Test openings, not ads
If the opening is where the outcome is decided, the opening is the unit worth testing. Four variants of the same ad with four different first two seconds will teach you more, faster, than four unrelated concepts — because everything downstream is held constant, so the difference is attributable.
Knack reads hook and hold rate together rather than ranking on either alone, and tags what the opening actually is — question, bold claim, POV, curiosity gap, relatable moment, direct address — from the creative itself. That's what makes the Patterns view able to answer "which kind of opening works on this account", which is the question a brief needs answered. On upper-funnel campaigns hook and hold are what Knack grades on, rather than marking an awareness ad down for ROAS it was never built to produce.
The thing worth un-learning
"Stop the scroll" has been the operating instruction for a decade, and it optimises for one metric — the view — while saying nothing about whether the view was useful. The attention research points somewhere more demanding: the opening has to hold someone and tell them who it's from, in less time than it takes to read this sentence.
Sources
- VCCP Media and Dr Karen Nelson-Field / Amplified (May 2025). Hacking the Attention Economy. 20,000+ views of 72 digital video ads, each tested against a "bad twin" with brand elements removed.
- Facebook / Nielsen Brand Effect analysis of 173 studies, reported by MarTech. Note the stated selection of positive-result campaigns.