Formats & the Matrix: 1 idea → 50 assets

All week you built the layers one at a time. Today you multiply them — on purpose — into the raw-material factory that feeds the entire learning loop.

1The last layer: formats want a native shape

This week walked down the diagram one layer at a time. Day 6 was For What — the Big Idea, the concept that can spawn many ads. Day 7 was For Whom — the persona you encode into the creative so the right human self-selects. Day 8 was the Message — the angle that lands for that persona. Day 9 was the Visuals — the hi-fi vs lo-fi fork. Today is the fifth and final layer: Format, the physical shape the asset takes — and then the assembly of all five into a system.

Format is not a cosmetic afterthought. Each Meta placement is engineered for a different consumption posture, and an asset built in the wrong shape gets punished by the auction before your message ever has a chance. The format axis of the genome (axis 7, from Day 4) covers the format type and its aspect ratio:

• Reels / Stories video — 9:16 full-screen, sound-on, vertical video. This is where lo-fi UGC lives and where the most volume now flows. Built for the thumb at full attention.
• Feed video — 4:5 or 1:1. The workhorse rectangle. Sound is often off, so the first frame and on-screen text carry the load.
• Static image — 4:5 or 1:1. A single frame that must win the 3-second war (Day 3) in one shot, with no hold-rate to lean on — so the first frame and its on-screen text do everything.
• Carousel — 1:1 swipeable cards. Ideal for sequenced argument: problem → solution → proof → offer, or three personas in three cards.

The rule: build the asset in the shape of the placement you want to win, then let resizing fill the rest. A 9:16 Reel cropped to a 1:1 Feed square loses almost half its height — usually exactly where the hook text sits. Native-shape-first, reformat-second.

2The multiplication: where 50 assets actually come from

Here is the move that defines a creative system rather than a pile of ads. You do not write fifty briefs. You write one concept and then multiply it through the layers you already built this week. Each layer is an independent knob — change one and you get a genuinely different, learnable variant.

Take a worked example. A sleep-supplement brand has one concept: "the off-switch for a racing brain." Now fan it out:

• 1 concept — the off-switch.
• × 3 personas — the burnt-out new parent, the shift-worker, the anxious high-performer.
• × 3 angles — pain-agitate-solve, before/after, social proof (from the angle library, Day 8).
• × 2 treatments — lo-fi UGC and hi-fi studio (Day 9).
• × 3 formats — 9:16 Reel, 4:5 Feed video, 1:1 carousel.

Multiply it out: 1 × 3 × 3 × 2 × 3 = 54 tagged assets from a single creative thesis — that's the "50" in this lesson's title; the exact count depends on how many values you give each knob. That is the combinatorial explosion — done deliberately, not by accident. And critically, every one of those 54 is born with its full genome tag (Day 4): a known concept, persona, angle, treatment and format. When the loop reads the results in Week 4, it can attribute the win to a component, because every component was labelled at birth.

3Don't make all 54: prioritise the highest-hypothesis cells

The matrix is a generator, not a mandate. The trap on the other side of "make more volume" is making junk volume — 54 assets where 40 of them test nothing you actually have a reason to believe. Combinatorics explode fast: add one more angle and one more format and you are at 96 cells. You cannot fund, ship or learn from all of them, and most are noise.

So the matrix becomes a prioritisation grid, not a production order. For each cell, ask: do I have a real hypothesis that this combination wins? The new-parent persona paired with a lo-fi UGC testimonial in a sound-on 9:16 Reel is a strong, specific bet — build it. A high-performer persona in a polished hi-fi carousel might be a weak guess you have no reason to hold — skip it for now and bank it as an Open Hypothesis for later (you'll meet that bucket on Day 18). This is exactly the explore-vs-exploit discipline from Day 5, applied at the cell level: spend your test budget on the cells with the highest expected information, not on every mathematically-possible combination.

A practical floor from the field: you want 5 genuinely distinct concepts beating 50 near-duplicates. To be clear: the matrix runs per concept. In production you keep 3–5 live concepts, each fanned into only its highest-hypothesis cells — concept diversity and within-concept variation are two different knobs, and you need both. Each variant needs roughly €20–€50/day for 7–14 days to escape noise — a rough floor for typical low-ticket DTC; the real test is whether each cell can collect enough conversions over the window to read, so scale the daily floor with your CPA (as of mid-2026 — check your own numbers). At €30/day, 54 simultaneous cells is €1,620/day — absurd for a test. Twelve well-chosen cells at €30/day is €360/day and gives every cell a fair, interpretable read. Prioritisation is what turns a combinatorial explosion into a fundable experiment.

4Many creatives, few ad sets: how the matrix meets the buying structure

Where do these tagged assets go? Not into 54 separate ad sets. This is where the creative side rejoins the demand side you learned in the media-buying course, Day 3: consolidation — many creatives inside few, broad, consolidated ad sets. You hand Meta a deep bench of varied creatives in one structure and let its inner loop (the auction + Advantage+ Creative — you'll meet the two loops formally on Day 16) allocate budget among them and surface winners within this generation.

That is the elegant fit: the matrix is the raw-material factory that produces the varied bench; the consolidated ad set is the arena where Meta's inner loop picks among them. Your outer loop — Week 4 — then dissects those winners back down into the genome components the matrix is built from, and briefs the next, smarter batch. Day 10 is where you stop thinking in "ads" and start thinking in a tagged inventory of controlled variants. Everything in Weeks 3 and 4 operates on this inventory.

Module check · Week 2

Five quick questions to lock in this module. Tap an answer to see if it lands.