The monthly brief · Issue 001
What 15 of our own ads say about the thing you're probably optimising
Issue 001: we scored 15 of our own launch concepts and the hook barely mattered. Plus what we're instrumenting on fatigue, and a teardown of our worst ad.
This is issue one, so a quick word on what this is. Once a month: what we're seeing fatigue, which angles are earning their spend, and one teardown worth stealing from. No pitch, no sequence behind it, one unsubscribe link in every send.
And a word on what it isn't. We're a new agency. We don't have a panel of forty accounts to average, and a brief that pretended otherwise would be worth exactly nothing to you. So this month you get the data we actually have — our own — and an honest map of what we can't see yet.
01 — The scoreboard
We scored 15 of our own ads. The hook explained almost none of the variance.
Before we launched, we ran our own 15 launch concepts through the same 0–100 scorer that scores client creative — five sub-scores (hook, clarity, distinctiveness, offer, brand fit) and a holistic overall that weights hook and clarity hardest. We published the results, losers included. The best scored 70. The worst scored 45.
Then we did something more useful than looking at the ranking: we checked which sub-score actually moved the overall. Across the 15, clarity correlated at r=+0.97 and offer at r=+0.91. Hook came in at r=+0.37, brand fit at +0.05, and distinctiveness at −0.11 — statistically, noise.
The reason is visible in the spreads. Hook scores ranged 60–68 across all 15. Distinctiveness ranged 62–70. Brand fit ranged 68–73. Those three were effectively constant, because they're a house style — once you've settled your art direction, every ad you make inherits it. Clarity ranged 40–71 and offer ranged 35–70. All the variance lived in what the ad actually said and what it actually offered.
The practical read: if your creative all comes out of one brand system, the hook is not your lever — it's your floor, and you've already set it. The thing separating your best ad from your worst is almost certainly whether a stranger can tell what you're offering in one read. That's a copy problem wearing a creative problem's clothes, which is why it survives so many rounds of new visuals.
Worth saying plainly: this is a scorer's opinion on one brand's 15 concepts, not a spend-weighted market result. See the limits at the bottom.
02 — Fatigue watch
What we're instrumenting, and why there's no number here yet.
The promise for this section is "what's fatiguing across DTC". We're not going to invent it. Here's the honest state: we have our own account and our first client accounts coming online, which is not enough to tell you anything about the category that you should act on.
What we're building toward, so you can hold us to it: for every account we run, we log the day each creative goes live, its score at launch, and the day its cost-per-result crosses back through its own first-week baseline. That gives an effective-life figure per creative that isn't a vibe and isn't borrowed from someone's content-marketing blog — it's the actual date a specific ad stopped paying, on an account we can see.
A caution while you wait for it, because this is where most of the published numbers in this category come from: the widely-repeated "creative fatigues in N days" figures are, in our reading, mostly recycled between marketing blogs with no methodology attached to them. If a fatigue number doesn't tell you the account count, the spend level and how fatigue was defined, it isn't a number — it's a mood. Your own account's data beats all of it, and you already have it.
Next issue will have the first real figures we can stand behind, or it will say why it doesn't.
03 — One teardown
Our worst ad scored 45. It was also one of the best-looking.
The teardown this month is our own, because it's the only creative we currently have the right to pull apart in public.
The concept was a retargeting ad. It scored hook 65 — right in the middle of the pack with everything else we made — and then clarity 40 and offer 35, for an overall of 45. Every other concept we shipped landed between 55 and 70. This one fell off the table on its own.
The scorer's note was blunter than we'd have been: visually striking, but the copy doesn't say anything. And it was right. The ad had a strong image, a confident layout, an on-brand palette — and a headline that gestured at a benefit without naming a deliverable, a price, a timeframe or an action. It looked like an ad for a company that had already explained itself somewhere else.
That's the failure worth stealing from, because it's the one that survives review meetings. Nobody kills an ad that looks good. It goes live, it underperforms, and the diagnosis becomes "creative fatigue" or "the audience is tired" — when the truth is the ad never communicated in the first place. A cheap test: show it to someone outside your company for three seconds and ask what you sell and what they'd get. If they can't answer, no amount of new visuals will fix it.
We kept it published on our own site with its 45 attached. Scores that only ever say "great job" aren't scores.
What this brief can’t tell you yet
- Sample size. 15 concepts, one brand — ours. Correlations from a set this small are directional, not conclusive.
- It's a scorer, not a market. These numbers say what a model trained on winning DTC creative thinks, judged against a published rubric. They are not spend-weighted results, and no claim here is that a 70 outperformed a 55 in an auction.
- No client data. We have no permissioned client results yet. When we do, they'll appear here named and numbered, with permission, or not at all.
- No fatigue figures. For the reason given above. We'd rather ship a section that says "not yet" than one that launders someone else's unsourced number into your strategy.
Next month
Want the next issuein your inbox?
One email a month, no pitch, one unsubscribe link — subscribe on the brief page. Or get this analysis pointed at your own account instead, free, in 24 hours.
2 minutes · findings in 24 hours · no call required