Grooming Vacuum Creative Teardown
Film the hair, not the machine: what 13 paid ads and 13 organic TikToks share, turned into a shoot and test plan.
- Client
- Manzanita Pet Co. (own brand)
- Role
- Research lead
- Tools
- Meta Ads Library, ffmpeg, local Whisper transcription, perceptual frame hashing, coded CSV dataset, AI-assisted synthesis
- Period
- Oct 2026
- Status
- filmed, in editing
At a glanceThe research has moved from analysis to filming to editing: the shoot is done and the first planned videos are being edited. No performance results are claimed until testing is complete.
Context & objective
Objective
Find repeatable creative patterns before filming — not to copy competitors, but to learn the structure this category uses and where there is room to stand out.
Challenge
Paid and organic videos follow different rules, and none of the files carried performance data. The analysis had to be strict about what the evidence can and cannot say.
Limits, stated up front: no performance data (ad longevity is a hint, not proof of profit); 9 of the 13 Meta ads come from one advertiser; no TikTok metrics; hooks are paraphrased and no ad copy is reproduced.
Creative strategy
Key findings
- The hair is the proof. A canister full of hair appears in 22 of 26 videos: all 13 TikToks and 9 of 13 Meta ads.
- Product first. The product or an attachment is on screen within 1 second in 20 of 26 videos.
- Same footage, two wrappers. The real split between platforms is voiceover, subtitles, graphics, CTA and end card, so one raw library can feed both.
- Savings travels. Groomer-cost savings appears in 10 of 26 videos and is the only primary hook type equally common on both platforms.
- Test hooks, not bodies. The dominant advertiser ran two exact hook-swap pairs with identical body, clips and clip order.
- Whitespace to own: explicit before/after (1 of 26), dog-POV narration (1 of 26), talking-head testimonials (about 2 of 26).
Meta vs TikTok at a glance
| Dimension | Meta ads (13) | TikTok organic (13) |
|---|---|---|
| Voiceover | 12/13 | 7/13 |
| Running subtitles | 10/13 | 0/13 |
| Product visible within 1 s | 8/13 | 12/13 |
| Explicit CTA | 7/13 | 2/13 |
| Median length | 55 s | 40 s |
Production process
Method
- 26 videos: 13 from the Meta Ads Library, 13 organic TikToks.
- Frames sampled at about 2 fps, with the first 3.5 seconds of every video reviewed frame by frame.
- Automated cut detection, and local transcription with word timestamps for speaking rate and coverage.
- Perceptual frame hashing to find footage shared or re-ordered across videos.
- One coded table (29 columns per video), followed by AI-assisted review and synthesis.
Deliverables & contribution
Deliverables
- A ranked list of 10 variables to A/B test, from hook line (first) to clip order (last).
- A 1–2 day shoot blueprint: about 130 clips tagged into 9 slot types, with shot list, specs and animal-welfare rules.
- Six video template specifications.
- A first wave of 10 videos: 3 hook variants on one body, angle tests and native organic formats, then test waves that change one variable at a time.
- Claims guardrails: no unsubstantiated health claims, decibels only after measuring, savings computed from real prices.
Contribution
I designed and directed an AI-assisted creative-analysis pipeline. I defined the research objective, selected the 26-video sample, decided what to compare, and directed the analysis and interpretation. Then I turned the findings into filming frameworks and a production plan.
Current status
The research has moved from analysis to filming to editing: the shoot is done and the first planned videos are being edited. No performance results are claimed until testing is complete.
Learnings
Creative research should translate directly into what gets filmed. The strongest pattern was showing proof, especially removed hair, rather than the machine. Paid and organic content can share the same raw footage while changing hooks, voiceover, subtitles and CTAs per platform.