Case study: make-it-virale
A short that flopped, three hooks as trial reels, a winner at 17,860 views: 4x the median and 3.8x the original.
The test that worked: /make-it-virale run on the /outliers short of
@melvynxdev. Same skill, same facts, same CTA. Only the hook changed, and the
video went from 4,711 to 17,860 views.
The first 8 seconds of each version (muted, click to turn the sound on).
The result
| Winner | 17,860 views, 270 shares, 950 saves |
| Against the original | 3.8x |
| Against the account median | 4x (median about 4,500 views, feed reels Oct 8 to 20) |
| Hooks tested | 3, all scored 8/10 on paper |
The Instagram screenshots, as they show in the app:




Only one is kept: the outlier goes to the profile, the other two stay in trial and your followers never see them.
1. The flop
The v1 went out in the feed on Oct 18 with this hook: « Ce skill cheaté m'a permis de passer de 5 000 à 51 000 vues sur mon compte » ("This cheat skill took me from 5,000 to 51,000 views"). At 24 h: 64 % skip rate, 9 s average watch, the worst of his last 12 reels.
The step 1 diagnosis:
- The hook talks about the creator ("took me"), not the viewer.
- Claude only shows up at 7 s. "Cheat skill" says neither what nor for whom.
- Frame 0 shows a small view count.
- The concept sells views, an angle that tops out at 1-2x in his niche.
The subject was good, the packaging was bad: that is the case make-it-virale
is for.
2. The winners on the same subject
Step 2 searches, through Treg and the inspirations, for reels that broke out on this exact subject (3x or more their own account's median):
- The "Claude spies on your competitors" angle: @lazymillionaire at 14.9x, @jayantcreates.ai at 60.2x.
- His own mould at 62x: « Tu peux maintenant transformer Claude en [métier] pour ton compte Instagram » ("You can now turn Claude into a [job] for your Instagram account").
- His
/stealshort: « Tu peux maintenant voler le style de montage de n'importe qui » ("You can now steal anyone's editing style").
3. Three hooks, three angles
| version | hook | mould |
|---|---|---|
| v2a | « Tu peux maintenant voler les vidéos qui cartonnent dans ta niche avec un seul skill Claude. » ("You can now steal the videos blowing up in your niche with one Claude skill.") | "You can now steal" (steal) |
| v2b | « Voilà comment tu transformes Claude en stratège qui espionne toute ta niche sur Instagram. » ("Here's how you turn Claude into a strategist that spies on your whole niche.") | remake of @jayantcreates.ai |
| v2c | « Tu peux maintenant transformer Claude en espion qui surveille tous tes concurrents sur Instagram. » ("You can now turn Claude into a spy that watches all your competitors.") | Claude as [job] + spy |
All three scripts passed the script rules and scored 8/10 on the hook. On paper, there was no way to tell them apart.
4. The trial reel test
The three versions went out as Instagram trial reels on Oct 19, two hours apart. A trial reel is only shown to non-followers: followers never see the same video three times, and each hook is judged by a cold audience.
5. The numbers
Measured on Oct 21 in Instagram insights.
| version | hook | views | reach | avg watch | shares | saves |
|---|---|---|---|---|---|---|
| v1 (feed) | « Ce skill cheaté… » | 4,711 | 3,555 | 15.2 s | 50 | 160 |
| v2a (trial) | « Tu peux maintenant voler les vidéos qui cartonnent… » | 17,860 | 14,054 | 10.1 s | 270 | 950 |
| v2c (trial) | « …Claude en espion » | 1,691 | 1,450 | 7.4 s | 16 | 61 |
| v2b (trial) | « …Claude en stratège » | 308 | 277 | 19.0 s | 0 | 5 |
What we learned
- The hook decides everything. Same video, same facts: from 308 to 17,860 views, a 58x gap between the worst and the best trial.
- Paper is not enough. Three hooks at 8/10, three unrelated results. Only the trial told them apart, which is why the skill tests three.
- The winner brings engagement too: 6x the saves and 5x the shares of the original, not just reach.
- What won: "You can now" + a strong verb ("steal") + the object + Claude in the first sentence, over a real proof at frame 0 (the 584K-view reel it copies), not a promise of views.
- Watch time does not predict the winner. v2b had the best average watch (19 s) and the worst reach: it held the few viewers it got but hooked almost none.
What's next
Step 6: the winner is shared to the profile from the Instagram app, then goes
out as a YouTube Short and a TikTok with /publish. The other two stay trials.
To run the same test on one of your shorts, see make-it-virale.