The 3-second rule is shorthand, not a law. TikTok’s creative guidance for performance ads says to introduce your content proposition in the first 3 seconds and to prioritize your hook in the first 6. YouTube changed how Shorts views are counted on 31 March 2025: a view now registers each time a Short starts or replays, with no minimum watch time, and the older measure lives on as “engaged views” (YouTube Help). Between them, those two facts describe the problem. Getting someone to start is cheap. A hook is whatever turns a start into a viewer who stays.
One limit up front. Retention curves belong to the account owner, so from the outside you cannot see who left at second 2. You can see what the opening contains and how the video performed relative to its creator’s norm. Every pattern below is a hypothesis to test in your niche, and the testing section below shows how.
Three channels carry every hook#
A hook is built from up to three channels, and the strongest hooks align at least two of them.
- Visual. The first frame and the motion in the first second.
- Verbal. The first spoken line, or the first line of a voiceover.
- Text. The on-screen words, which also carry the hook for viewers watching without sound.
When you review a video, write down what each channel says. If they say three different things, the opening is probably doing less than it looks like.
The nine patterns#
| Pattern | What the first 3 seconds contain | Fits | Common failure |
|---|---|---|---|
| Result-first | The finished outcome on screen before any explanation | Visible transformations: skincare, cleaning, organizing, styling | Result looks edited or unrelated to the product |
| Problem call-out | A direct name for a pain the viewer has | Problem-aware buyers, utility products | Pain is generic, so nobody feels addressed |
| Pattern interrupt | An unexpected object, motion or cut that breaks the scroll | Low-interest categories, crowded niches | Interrupt has no link to the offer |
| Open loop | A promise of a payoff withheld until later | Stories, experiments, reviews | The payoff is weaker than the setup |
| Contrarian claim | A statement that disagrees with common advice | Education, expert-led products | Claim cannot be defended in the next 5 seconds |
| In-use demo | The product working, no talking, sound optional | Gadgets, kitchen, tools, beauty tools | Demo needs context a silent viewer lacks |
| POV scenario | A situation the viewer can step into (“POV: …”) | Lifestyle, comedy-adjacent brands | Scenario reads as a skit, not as a product reason |
| Numbered promise | A count that sets a finish line (“3 things I stopped buying”) | Roundups, comparisons, tips | Numbers pad the video; item 1 is already weak |
| Credential line | A role or experience that earns attention (“I’m a nurse…”) | Health-adjacent, finance, B2B | Credential is unverifiable or overstated |
Trend formats sit on top of these. A trending sound or template is a wrapper, and the pattern underneath is still one of the nine.
Three illustrative examples#
For a heel balm, written as briefs rather than as findings:
- Result-first. Frame 1 shows a smooth heel next to the tub. Text reads “14 days.” The voice says, “This is the only thing I changed.”
- Problem call-out. Frame 1 shows socks catching on rough skin. The voice says, “If your socks snag on your heels, watch this.”
- Contrarian claim. Frame 1 shows a pumice stone going into a bin. The voice says, “Stop filing your heels.”
Each one promises something different, so each needs a different proof beat in seconds 3 to 10.
How to find which patterns win in your niche#
You need a set of videos, a tagging pass and a comparison.
- Pull 80 to 200 relevant videos from one niche and one time window across TikTok, Reels and Shorts.
- Tag each opening with one primary pattern, plus the channel notes above. Resolve disagreements between taggers with a written rule, such as “the pattern is whatever the viewer experiences first.”
- Normalize performance. Divide each video’s views by its creator’s median views. Rank by that ratio, not by raw views.
- Compare shares. For each pattern, compare its share of the top quartile with its share of the whole set.
Lift is the top-quartile share divided by the overall share. An illustrative result from 80 tagged videos, where the top quartile holds 20:
| Pattern | Overall | Top quartile | Lift |
|---|---|---|---|
| Result-first | 20 of 80 (25%) | 9 of 20 (45%) | 1.8 |
| Problem call-out | 24 of 80 (30%) | 5 of 20 (25%) | 0.8 |
| Pattern interrupt | 8 of 80 (10%) | 4 of 20 (20%) | 2.0 |
| Open loop | 12 of 80 (15%) | 1 of 20 (5%) | 0.3 |
| Contrarian claim | 6 of 80 (7.5%) | 1 of 20 (5%) | 0.7 |
The other four patterns make up the remaining 10 videos, none of them in the top quartile.
Read this table with care. Pattern interrupt has the highest lift but only 4 videos behind it, so it is a lead to test. Result-first has 9 behind it, so it is a stronger bet. Problem call-out is the most common pattern and sits below average in the top quartile, which suggests it is the default people reach for, not the one that wins.
Check the payoff before you copy a pattern#
Open every top-ranked video again and look at seconds 3 to 10. Does the claim from the first three seconds get proved? A result-first hook with no visible product use afterwards is a promise without evidence. A contrarian claim with no reason attached is an argument starter. Check whether the winners in your set pair the opening with a payoff inside the first ten seconds, and brief that pairing, not the opening alone.
Two other checks are cheap:
- Sound off. Mute the video. If the opening still communicates, the text and visual channels are working.
- Native feel. TikTok’s guidance recommends a DIY, not overly polished look featuring creators, employees or customers. If your reference videos look like ads and the winners look like posts, copy the winners.
Matching a pattern to a product (starting hypotheses)#
| If your product… | Test first | Then |
|---|---|---|
| Shows a visible change | Result-first | In-use demo |
| Solves a familiar annoyance | Problem call-out | Numbered promise |
| Needs trust before purchase | Credential line | Contrarian claim |
| Competes in a crowded category | Pattern interrupt | Open loop |
These are places to start, not rules. Your own tagged set will overrule them.
Running the taxonomy in ContHunt#
Deep Search can tag every video it analyzes against a fixed label list. Give it the nine patterns as the allowed answers, so counts stay comparable:
For each video, label the opening hook as one of: result-first, problem call-out, pattern interrupt, open loop, contrarian claim, in-use demo, POV scenario, numbered promise, credential line. Note the first frame and the spoken line. Then compare the pattern mix of the top-performing videos with the rest of the set.
ContHunt returns the analyzed set with the labels and links to the videos behind each count. Its scores come from analyzing the content and its fit to your question. They are not measured viewer retention, so keep the creator’s own analytics as the final check.
Once you know which pattern leads your niche, put three variants of it into your next creator brief. Get started with a Deep Search on your niche and see how your own hook mix compares.
