Pull the library
Extract the patterns
Cluster the angles
Write the brief
Ship the concepts
Context & problem
Creative testing dies on the brief. Most DTC teams and agencies know they should mine competitor ads, but it turns into weeks of scrolling with nothing to show for it — a folder of screenshots and a gut feeling dressed up as strategy.
By the time a brief reaches the designers, it is usually the same recycled hooks tested against each other, with no signal about which angles competitors have already validated with real spend. You end up paying to learn things the market already knows.
This is the sprint I built and ran for a DTC performance agency that needed briefs their designers could act on without another week of scrolling. I packaged it as a fixed-scope sprint because it is a research problem with a clear finish line, not an open-ended retainer. Five days in, you have testable creative, not more research.
How it works
The diagram above shows the flow; here is what each step does. Deterministic stays deterministic, and an agent only shows up where judgement, language, or synthesis is actually needed.
- 01Deterministic
Pull the library
I pull competitor ads across the category and adjacent ones from public ad libraries. It is a defined collection step — same sources, same fields captured every time.
- 02Agent
Extract the patterns
A model reads each ad and tags the hook, offer, format, opening, and proof points. This is language and interpretation work, so it is where the AI earns its place.
- 03Agent
Cluster the angles
The tagged ads get clustered into repeatable angles, prioritized by how often they show up and how strong the signal is. Deciding what counts as the same angle is the hard part, and it happens in language.
- 04Agent
Write the brief
The findings become a one-page creative strategy brief written for designers, not founders — concrete enough to hand off. I review and tighten it before it ships.
- 05Deterministic
Ship the concepts
Three test-ready ad concepts, each with a landing-page outline, packaged for design. Fixed deliverable, fixed scope.
Checkpoints & logging
The sprint has one built-in checkpoint: a 30-minute kickoff to lock the category and the buyer before any research runs. Point it at the wrong competitors and you get a beautiful brief for the wrong fight, so that alignment happens first, on a call, in writing.
Every ad that goes into the analysis is logged with its source and tags, so the brief is traceable — you can see which specific ads a recommended angle came from instead of trusting a summary. None of the output is a black box you have to take on faith.
Stack — and why
The orchestration runs on n8n so the collection and tagging steps are inspectable — when a tag looks wrong I can see the exact ad and prompt that produced it. The pattern extraction and clustering lean on a language model because that is genuinely interpretive work; a keyword filter would miss the angle behind the words. Everything else — the collection, the sorting, the packaging — stays deterministic so the deliverable is repeatable sprint to sprint.
Results
Per sprint you get roughly 20 competitor ads analysed and tagged, around 7 repeatable creative angles, 3 test-ready concepts, and 1 one-page brief. Fixed scope, fixed price, five working days.
Run for a DTC performance agency against their category, the output went straight to their designers as a brief instead of a folder of screenshots. The point is not volume of research — it is that what lands in front of your designers is grounded in what the category has already validated, not another round of guessing.
Want this run for your category?
It is a fixed-scope, fixed-price sprint — five working days from kickoff to a brief and three concepts. Book a call and I will send back a one-page sprint plan for your category within 24 hours.