MoonDesk Scaling Method
We define who you’re selling to, find the angles that move each persona, and test them until the data tells us what scales.
The MoonDesk Scaling Method is how we plan and test ads for fashion brands. We define the customer personas who buy your product for different reasons, match each one with the angles most likely to persuade them, turn the best pairs into genuinely different concepts, then test, read and iterate every week on what the results show.
We build every account around persona × angle testing: we define the customer personas, match each one with the angles most likely to persuade them, and test those concepts in Meta’s Andromeda-era delivery system, iterating every week on what the data shows.
Why it matters now
Ad platforms now decide who sees your ads, and they decide largely from the ads themselves. That moves the real targeting work into the creative.
Andromeda is the ad retrieval engine Meta introduced in late 2024. It narrows tens of millions of ads to the few thousand each person could see, before ranking picks one. Meta’s engineers put it plainly: “Increased ad diversity can improve people’s experience with ads and drive better advertiser outcomes.”
In Advantage+ sales campaigns, Meta finds the audience, and its own best-practice list tells advertisers to “provide a wide variety of diverse creative assets.” A narrow interest stack matters far less than it did five years ago.
Meta also warns that “too many ads can result in worse performance.” Fifty near-identical ads split your budget without giving the system anything new to work with. What helps is difference — a different persona, message, visual world and format.
So the question isn’t “how many ads?” It’s “which people, and which reasons to buy?” That’s what the method answers.
The six steps
The three words our clients know, each turned into concrete, repeatable steps. The loop restarts every one to two weeks.
We study your real buyers — reviews, post-purchase surveys, comments, customer interviews, Shopify and GA4 cohorts — and define the three to six personas who buy the same product for different reasons.
Output: Persona sheet: who they are, what triggers the purchase, their objection, the identity they want, the words they use.
For each persona we list the angles most likely to persuade them — fit and comfort, styling, quality, social proof, value, scarcity, gifting, founder story — and rank them by expected strength.
Output: Persona × angle matrix with a written hypothesis in every cell.
We turn the strongest cells into genuinely different concepts: a different persona, message, visual world and format. Not ten edits of one ad.
Output: Concept briefs, hooks, scripts, statics, UGC and video.
We launch the concepts in one consistent structure — broad targeting or Advantage+, controlled budgets, enough spend per concept to read — with the persona and angle tagged in every ad name.
Output: Test log: hypothesis → concept → spend → result.
We read results by persona and angle, not just by ad: which personas convert, which angles win, which hooks hold attention — net of returns.
Output: A learning, e.g. “Gift buyers + fit confidence wins on Meta, loses on TikTok.”
We give winning pairs new executions, cut the losers and write the next hypotheses. Every round adds to a creative learning library that compounds.
Output: The next round of the matrix.
Step 6 feeds step 2: every round of results writes the next round of hypotheses.
The persona × angle matrix
An illustrative example for a fictional linen dress brand. Three personas and three angles give nine hypotheses; each cell is a bet on why that persona would buy. Pick a cell to see the concept it becomes.
The occasion shopper × Wear it three waysWinner
The occasion shopper × Fit you can trustIn test
The occasion shopper × Made to lastCut
The capsule minimalist × Wear it three waysNext round
The capsule minimalist × Fit you can trustIn test
The capsule minimalist × Made to lastWinner
The gift buyer × Wear it three waysCut
The gift buyer × Fit you can trustWinner
The gift buyer × Made to lastIn test
Fictional brand, illustrative statuses. Real matrices come from your customers’ own words.
A test round, week by week
Reviews, surveys, comments and store data become the persona sheet and the matrix. We also check tracking and your product catalog.
We pick the strongest cells and brief genuinely different concepts — then shoot, edit or source them with creators.
The first round goes live in a clean structure, every ad tagged with its persona and angle.
We read by persona and angle, net of returns. Winners get new executions; losers get cut; the next round gets written.
From here the loop runs every week or two, and the learning library grows with every round.
What you get
Three to six personas in your customers’ own words: trigger, objection, desired identity.
Every cell a written hypothesis, ranked, with its status.
Persona, angle, hook options, format and must-haves for every concept we produce.
Hypothesis → concept → spend → result, for every test we run.
Which personas convert, which angles win, on which platform — net of returns — and what we test next.
One method, five platforms
Personas and angles belong to your customers, not to a platform. What changes is the format and the job each platform does.
The method in practice
MESSAGE Footwear started working with us in February 2026. We scaled their Meta and Google ads, then activated Pinterest and TikTok — every platform running on the same persona × angle tests.
Source: MESSAGE Footwear’s Shopify gross sales, all channels, February–September 2026 vs. February–September 2025. Gross sales include every channel, not only paid ads.
Bring your product and your numbers. In 45 minutes, Hugo will sketch your first personas and angles with you — and show you where we’d start testing.
Most fashion brands need three to six personas. We start testing with the two or three that buy the most, so each one gets enough budget to read, and add the others once the first round shows which reasons to buy are strongest.
Count concepts, not ads. Meta publishes no official number and warns that too many ads can hurt performance. What matters is that each round adds genuinely different concepts — new persona, angle or format — and that your budget lets each one get a fair read.
At least a week. Meta says an ad set usually needs about 50 results in the week after its last significant edit to exit the learning phase, so we read on a weekly rhythm and wait for enough purchases before calling a winner.
Yes. Personas and angles describe your customers, so they carry across platforms; only the format changes. On TikTok an angle becomes a creator video, on Pinterest a pin built for a search or an occasion. The same matrix often wins on one platform and loses on another — that’s a learning, too.
An A/B test compares two versions of one thing — a headline, an image — to find the better one. Persona × angle testing compares hypotheses about who buys and why, using concepts that differ on purpose, and reads the results by persona and angle. It tells you what to make next, not only which ad won.
Enough for each concept to reach a meaningful number of purchases in a week or two. Work backwards from your cost per purchase: Meta’s guideline of about 50 results a week per ad set means that at $40 per purchase, an ad set needs roughly $2,000 a week to settle. If that’s out of reach, we test fewer concepts at a time rather than starving all of them. Our ROAS calculator shows your break-even point first.
Bold, story-led paid media for fashion brands ready to grow — clear messaging, outstanding creative, and a method that turns every test into the next win.
*A 45-minute strategy session with Hugo to see if we’re the right fit for your brand.