Most ad creative gets made the way it has always been made: a brief, a shoot, a cut, a launch. Then everyone goes silent and waits to see what happens. By the time the data comes back, the hooks are dead. We changed the unit. The unit is not the ad. The unit is the loop.
01 The loop, not the ad
On TikTok, hooks fatigue every 5 to 7 days. On Meta, the cycle is 2 to 4 weeks. After a week of running the same creative, CTR starts to slide and CPA starts to climb. Those are not opinions, they are the working assumptions every paid team is now stuck with.
The implication is uncomfortable for traditional agencies: a single ad, no matter how well crafted, is a depreciating asset that hits zero inside a month. The asset that does not depreciate is the system that produces the next ad. That is what we sell.
02 Phase one: brief
Every cycle starts the same way. We open Meta Ads Library and TikTok Creative Center, pull every ad your three closest competitors are running, and document the hooks, formats, and CTAs. We do the same for the brand we think is winning your category that you are not paying attention to.
Then we generate. Five hooks per concept, three concepts per cycle. The hooks are tagged by category: problem-solution, POV, transformation, social proof, curiosity gap, competitor switch, listicle. We write the script for each one in the creator's voice, not ours.
What you get at end of phase one
- A competitor ad audit, sourced and dated
- 15 hook variants tagged by category and platform
- 3 base scripts with hook + CTA swap matrix
- A creator brief ready to ship
03 Phase two: cast
The brief goes to the right creator from our managed network, not whoever is available. The match is on three vectors: vertical familiarity, on-camera energy, and the format the brief calls for. A founder-led B2B SaaS hook is not the same casting call as a fitness app POV.
Creators shoot on phone, in their own environment, in 72 hours. Raw clips come back via Drive. We review every clip within 24 hours of receipt. If the hook delivery is off or the audio is dirty, we flag the reshoot before it leaves the creator's calendar for the week.
The casting calls we run
- Talking head, founder-style, problem-first
- Talking head, peer review, comparison
- POV, app-engagement, mascot or absurdist
- Screen recording with voiceover, demo or workflow
- Day-in-the-life, authority figure, soft sell
04 Phase three: cut
Each base video becomes 2 to 3 ad-ready variations. The same body, three different first seconds. The same body, two different CTAs. The same body, three different lengths. This is where volume comes from, and where most teams stop short.
Captions are baked, not auto-generated. Aspect ratios are platform-specific (9:16 for TikTok and Reels, 4:5 for Meta Feed, never square). Audio is checked for sound-off legibility. Pacing is tight: long enough to land the CTA, short enough to keep hold rate up.
Hook swap
Same body, three first-seconds. The single highest-ROI test we run.
CTA swap
Soft close ("link in bio") vs direct close ("download now"), tested in pairs.
Length cut
15s, 25s, and 45s edits of the same script. Each platform prefers a different one.
Format swap
Talking head vs screen recording vs text-first edit, where the script supports it.
05 Phase four: iterate
Every cycle ends with a read. After seven days live, you send us back the spend, impressions, hold rate, CTR, and CPA per variation. We rank winners against the control. The winner gets fed into next week's brief as a constraint: this is what is working, build the next variation against it.
This is where the moat lives. The first cycle is just creative. The fourth cycle is creative informed by three weeks of your account's specific data, plus pattern data from every other account we run. The compounding is not theoretical. It is the only reason we charge a retainer instead of project rates.
06 The week, in days
Competitor pull, hook generation, script drafting, brief sent to creators by 6pm.
Creator match. Product shipped if physical. Calendar locked for shoot day.
Raw footage uploaded by EOD. We review same evening.
Hook swaps, CTA swaps, length cuts, captions, platform exports. Delivered Sun afternoon.
Variations into your ad manager. Test budget allocated. We give you the audience pairings and recommended splits.
Hold rate, CTR, CPA per variation. Winner declared. Loser killed. Constraint set for next week's brief.
07 Where AI sits in the stack
We use AI tools where they save time without costing performance. Hook generation against a curated prompt library. Script drafts that get rewritten by a human. Editing accelerators in CapCut. Avatar tools for testing a hook before we book a real creator.
We do not use AI to replace the creator. Real human creators still beat synthetic content on hold rate, especially in testimonial and personal-story formats where audiences are most sensitive to anything that feels artificial. The gap is large enough that no production cost saving makes up for it. We are honest about that, in both directions.
08 What we do not do
- We do not build brand films. There are agencies for that. We are not one.
- We do not run your media buying. We hand off creative to your team or your buyer.
- We do not take on accounts under $10k a month in paid spend. The math does not work for either of us.
- We do not sign annual contracts. 30-day rolling. Pay before each cycle.
- We do not publish client cases without explicit permission. Most growth teams want their creative strategy unpublished.
09 The thing nobody writes down
Most agencies will not say this part because it makes the retainer look smaller. Here it is: the first month is the proof. If we have not moved your CAC inside 30 days, you walk and we have the conversation about why. We are not optimizing for retention through lock-in. We are optimizing for retention through the number.
That is the playbook. Twenty-six pages worth of operating detail compressed into one. If you read this and the loop maps to how your team thinks about creative, the next step is twenty minutes on a working call.