Why this studio exists, what it believes, and how the work actually gets made.
Most performance creative is bad on purpose. The economics of an agency say onboard as many brands as you can and touch each one as little as possible. Then ship enough volume that something works.
The ads look like that math.
So buyers see "performance creative" and assume it means cheap, fast, and disposable. They start asking for fewer ads, not better ones. They stop trying.
I started Miru Studio because I think this is fixable. A creative director can run weekly cycles against a real ad account, hold the brand bar, and ship the volume an in-house team cannot. The work that requires taste stays with the creative director. The work that requires throughput moves to software. That is the whole thesis.
Five seats, one chair.
On every account I run, I carry five roles: designer, filmmaker, photographer, analyst, and the production system that ties them together. Five seats a small agency staffs separately and then watches fight each other for direction. The designer wants the pretty thing. The analyst wants the big number. The filmmaker wants the beautiful shot. They each work toward their own goal and the work is the casualty.
Running all five from one chair is the moat. The designer's instinct doesn't override the analyst's read. The analyst's read doesn't override the brand bar. The AI layer doesn't override the work. They all serve the same number on the same account.
Every ad gets made by hand.
AI reads the data and fills production gaps no one would shoot for. It does not write the ad. It does not pick the hook. It does not get a vote on taste. There is a category of tool right now that takes one photo and spits out a thousand variants. That is not what this is. That is noise with a higher ceiling than handmade work, until the moment the brand has to live with the shelf full of it.
I use AI the way a designer uses Figma. As a layer underneath the work, not the work itself.
What I believe.
Volume without taste is noise. Volume with taste is leverage. The best ads are made by people who can shoot, edit, design, and read the data, all on the same account, in the same week. A brand number and a performance number reward each other when the work is good. This studio is one creative director, one weekly cycle per brand, and a pipeline built to compound.
The shoot-edit-design-data combination didn't come from a marketing thesis. It came from doing each of those jobs separately for fifteen years. First paid app on Roku in 2009, then design lead at iSpot through the Goldman acquisition, then global brand films at Matador with a festival-awarded documentary out of Japan, then a $1M education vertical at Moment with a two-person team. Different jobs, same instincts. The studio is what those instincts look like aimed at one account.
I cap the roster on purpose. Selectivity is part of what you're paying for. I am not chasing scale.
The methodology
Performance creative is a search problem. You map the territory, narrow it with data, and double down on what's moving the account. Every brand runs the same five-step workflow, tuned to their category, footage library, and ad-account history.
Every engagement opens with a read of what's already running. Ad account, organic library, prior agency work, customer feedback, category context. What's working, what isn't, what shouldn't still be live, what the data already tells you. No assumptions before the audit.
Eight to twelve testable themes pulled from the audit and the category. Education, problem-solution, founder-led, podcast-style, customer testimony, ingredient breakdown, anti-claim, comparison. Each one a hypothesis about why someone stops scrolling. The map is the search space.
The library gets indexed first. What you have, you use. The gaps get closed three ways: repurposing existing footage, AI-generated material trained to your brand's aesthetic, and production shoots when authenticity is the hook. AI for filling, not for making. The brand approves every generation.
Hooks get written through brand-trained agents and scored against three filters: does it sound like the brand, does it earn the watch, does it set up the close. Most candidates die at this layer. The ones that survive get cut by hand.
Weekly cycle with the buyer. What carried spend, what fatigued, what's the next candidate off the back of a winner. Kill losers proactively, scale winners, pull two or three new candidates every month. The loop runs itself by month three.
The weekly cycle inside step five has a name: the Miru Loop. Library indexes the brand's footage. Rewrites turns winning hooks into five-plus angles for next week. Score grades every variant before it ships to the buyer. The Loop is what makes step five sustainable across a year.