Ad Modules & Copy Placement

The problem it solves

Somebody writes the long version of a campaign argument and hands it to a designer. The designer finds out the argument doesn't survive at 320×50, so it gets rewritten at the placement. Do that thirteen times and you have thirteen slightly different arguments, none of which were approved, and no way to tell whether a test result was about the message or about the rewrite.

The eight fragments

A module is eight pieces. Two are shared across the campaign; six are specific to the module. Each is authored at up to three lengths.

#	Fragment	The moment it creates	Tell that it's wrong
F1	Tension	Recognition. They see their own week. Slight discomfort, not alarm.	It's about the category instead of about them.
F2	Claim	We take a position. Confident, flat, declarative.	Hedged, or it's a benefit rather than a position.
F3	Turn	The implication lands. The small drop when something clicks.	It restates the claim instead of following from it.
F4	Hero proof	Evidence. The argument stops being opinion.	It needs a sentence of setup to make sense.
F5	Supporting proof	Depth. It wasn't a fluke. Accumulation.	It repeats the hero stat in different clothes.
F6	Stake	Consequence. Steady, never dramatized.	It's a threat, or so vague it fits anything.
F7	Position	We enter the argument. Earned certainty, quiet rather than loud.	A competitor could say it.
F8	Ask	The exit. Easy, almost casual.	It asks for more than the offer justifies.

F7 is the one that gets skipped and shouldn't. It's the only fragment that makes the ad yours rather than the category's. A module with no F7 is buying awareness for a topic, including, usually, for your competitors.

Three length tiers

Tier	Limit	Where it runs
L · Long	25 to 45 words	Post copy, video VO, landing page, email, script
S · Short	45 characters	Carousel cards, headlines, large display
C · Compressed	30 characters, must name the category	Small display, search headlines, anywhere one text element survives alone

Three tiers, not four. The 45-character ceiling comes from the tightest carousel card and clears a 70-character headline with room. The 30-character ceiling comes from a search headline field, which is usually the tightest thing in any media plan. Set both from your own plan before you start.

The compression ladder

Placements are recipes. They reference fragment IDs and tiers, never content, which means this table works for every campaign you ever run.

Placement	Recipe
Social post copy	F1-L + F2-L + F4-L + F3-L + Source-L
Social headline	F2-S
Carousel, five cards	F1-S / F2-S + F5-S / F4-S / F3-S or F6-S / F7-S + F8-S
Display, medium rectangle	F2-S headline + F4-S subhead
Display, leaderboard	F2-C
Display, mobile banner	F1-C or F2-C
Search headline / description / sitelink	F2-C / F4-L trimmed / F8-C
Email subject / preheader	F1-S / F4-S
Landing page hero / subhead	F2-L / F4-L
Video, 30 seconds	F1-L, F5-L walked as stages, F3-L, F7-L

Six rules

1. One fragment per text element. A card carrying a claim and a stat is two elements with two fragments. Merged sentences don't compress.
2. C tier names the category. Any placement where a single text element survives alone has to contain the category word. Display reaches people with no prior exposure and no reason to guess.
3. A fragment that can't reach C tier isn't a fragment. If the claim only works at 45 words, it's too complicated to run a campaign on. Finding that out during authoring is cheap. Finding it out during build is not.
4. Figures shorten, never paraphrase. The moment a number gets reworded it stops being citable. Every figure carries its hedge into every tier, or it drops out of that tier.
5. Fragments are versioned, placements aren't. When a figure changes or legal clears something new, edit one fragment and every placement inherits it.
6. Modules don't share fragments except F8 and Source. If two modules use the same hero proof, they're the same module wearing two headlines, and a test between them reads as inconclusive.

The intake session

Seven questions, roughly fifteen minutes, answered out loud rather than written. Voice is the point. The best short lines usually already exist in how someone says it before they start composing. Mining the transcript beats writing from notes.

1. Who is this for, and what's on their plate this week?
2. What's the most uncomfortable true thing we could say to them?
3. What's our answer? One sentence, no hedging.
4. If they remembered one number, which one, and what's it actually proving?
5. What does it cost them to be wrong, and why does that matter now?
6. If a competitor put their logo on everything you just said, what would be wrong about it?
7. Say the whole thing in one breath, like you're explaining it walking to a meeting.

Two follow-ups, used sparingly. When an answer comes out in marketing register: "say that again like you're telling a friend at a bar." When an answer to 4 or 6 comes back thin: "why that one?" Once, not repeatedly.

Questions 1 through 3 happen before anyone opens the stat deck. Under ten seconds means it isn't thought through; over three minutes means ask it again, shorter.

A worked module

Using a manufactured brand, Meridian, a hypothetical field service operations platform, with invented figures.

#	Tier	String
F1	L	Approvals happen wherever somebody happens to be standing. The decision gets made in a text message and the record gets made three weeks later, from memory.
F1	S	The record is written from memory.
F1	C	Field spend, unrecorded
F2	L	Losing control of spend is what happens when the decision and the record live in different places.
F2	S	Control is a workflow, not a habit.
F2	C	Field service control
F3	L	Which means the fix isn't more vigilance from people who are already stretched. It's putting the decision and the record in the same place.
F4	L	Teams reviewing line items in a shared workflow reject 14% of submitted work. Teams reviewing after the fact reject 2%.
F4	C	14% rejected vs 2%
F6	S	Small charges compound quietly.
F7	L	Anyone can store the record. The question is whether it exists at the moment somebody has to decide, which is a workflow problem rather than a storage one.
F7	S	The record has to arrive first.
F8	C	See the workflow

What it costs

This front-loads work into a phase that used to feel optional, and the first campaign will feel slower. Worth naming rather than hiding.

The two structural wins: a designer picks a fragment already written to size instead of rewriting to fit, and a changed statistic means editing one fragment instead of hand-patching thirteen assets. Baseline these on the next campaign so the second one has something to compare against.

• Time from brief to build-ready copy. Expect it to get worse on the first campaign and materially better by the third.
• Rewrite rate. Percentage of placements where copy changed after design started. Target is zero. Above ten percent means the C tier work got skipped.
• Review cycles per campaign. Legal reviewing a stat bank once beats reviewing thirteen assets.
• Percentage of tests with an explainable winner. The failure case is a result nobody can act on.
• Reuse rate across campaigns. The compounding benefit, and it only shows up after several cycles.