From 6.2M accidental plays to a repeatable narrative-first system.
A cosplay creator with viral hits she didn't understand and "vibes" posts that were quietly killing her reach. We took her own data and built the system she'd been running by accident.
The brief
Kirahri had viral hits — "It be like this" at 2.8M plays, "I uhhh went over budget" at 1.9M — and a pile of underperforming reels in the same week. The advice she'd been given was a mix of standard creator playbook ("post more," "use trending audio," "be authentic"). None of it explained why two reels she shot in the same outfit could differ by 2,800x in plays.
Her engagement-to-likes ratio was 5.04% — seven times higher than her largest competitors at 400K+ followers. The audience was already there. The system underneath the posting wasn't.
What the data actually said
We pulled every reel from the past 18 months and tagged each one across six content pillars (Convention, Behind-the-Scenes, Humor, Personal, Social, Showcase). The pattern surfaced in the first hour of analysis:
- Behind-the-scenes content averaged 11.4% engagement rate. Convention content averaged 9%. Humor averaged 10%. Together: 90%+ of total plays.
- Pure trend/meme posts (no personal angle) averaged 1,681 plays. Motivational posts averaged 1,644 plays. The biggest plays-killer in the entire feed was "advice" content.
- Caption format predicted reach. "I [did something specific]" hooks consistently outperformed vague captions by 100-2,800x. Same creator. Same outfit. Same audio in some cases. The first line of the caption was doing all the work.
- Audio that ADDED meaning beat audio that just played. "Just A Girl" by No Doubt with a Just-A-Girl-themed cosplay = 69K plays. The same energy creator with no audio-meaning match = under 2K.
What was built
- 26-document content system covering pillar mix (2 Tier 1 + 2 Tier 2 + 1 Tier 3 max per week), caption formulas that worked for her specific voice, audio-meaning rules, hook libraries by location, and a competitive scan of three peer accounts.
- Three pre-built reel concepts shot-by-shot, including framing, on-camera lines, expected hook, and the specific caption to use.
- Hook library by location — different hooks for at-con vs. at-home vs. in-car, so she could film what she had access to without forcing the wrong format.
- Mute test as a quality gate. If the reel reads identically on mute, the audio isn't earning its keep — change the audio or change the cut.
- Audio classification framework — when to use her own voice vs. franchise OST vs. trending audio vs. licensed song-as-the-joke.
- Testing framework + 30-day calendar. Pre-scripted week-one through week-four content, each post tagged with the hypothesis it was testing.
The lesson
Viral hits in a creator's feed are usually not flukes. They're the unconscious version of a system the creator already knows, articulated through whichever post happened to make it past their hesitation. The job of strategy is to find the unconscious system, name it, and make it the default — not to invent something new.
Most creators (and most local businesses) are running the same architecture: a few posts that worked, a lot that didn't, no theory of why. Surface the why, build a calendar around it, and the median post starts performing closer to the top than the bottom.