Service 07
On Meta and TikTok the ad decides who sees it. Audience settings barely matter; variety and volume do. Creative gets briefed from what the performance data says is missing, produced in enough genuinely different concepts to give the algorithm something to match, and read like any other experiment. We build that engine together, and it keeps running in your team.
Creative quality accounts for roughly half of what advertising sells and gets about a third of the budget. It’s the largest lever left in most paid accounts and the one most often handed to people with no view of the performance data, so it gets judged on taste in a review instead of on the number in the account. Meanwhile the platforms rebuilt themselves around it: a campaign with three ads never enters the auction, and ten near-identical hooks count as one.
Here the brief starts in the numbers, because the person writing it also reads the media. AI drafts the volume for cold-traffic testing; people make the winners worth scaling; creators carry the ones that need a face. Every concept gets a hypothesis, enough variants to isolate a variable, and a read at 48 hours on the early signals that actually predict, before anyone argues about ROAS on a sample of nine.
Ten to fifteen genuinely different concepts per campaign and new ones every month, in the formats the platforms distribute: short video first, statics and carousels where they still earn their place. Hooks, angles and formats, not colour swaps. Retired on fatigue, not on a calendar.
Generative tools for what they’re good at: dozens of variants for cold-traffic testing at a fraction of the cost. Humans for what they’re good at: the winners, the brand, anything a customer will look at closely. Nothing generated ships without a person approving it, because the platforms will happily rewrite your copy for you.
Real people for the categories where testimony is the purchase driver, briefed against the same hypotheses and whitelisted so the ad runs from their handle. The authenticity is the whole pitch; it can’t be generated yet, and the audience can tell.
The conversion assets that decide whether the traffic you paid for turns into a sale on Amazon: imagery, A+, Brand Store, video on the product page. Treated as capital with a book value, audited against live listings, fixed in priority order. See Amazon & Marketplaces.
Briefs, revisions and approvals across a portfolio, keeping one standard when several brands and studios are moving at once. Both chairs, maker and director, so the brief is written by someone who knows what it takes to produce.
Work that goes to a press and can’t be fixed after launch: packaging, direct mail, signage. And the screens: wireframes through finished mockups for sites and apps, then built, since the front end ships here too. See Web Development.
Which query the traffic arrived on, which objection the reviews keep raising, where the drop-off is, which hook stopped the thumb last month. A brief that starts from a mood board is guessing expensively.
One execution isn’t a test. Concepts come in variants designed to isolate a variable, so the winner tells you something you can use again. Your team learns the system so the volume keeps coming without me.
Thumb-stop and hold rate at 48 hours, not ROAS on a sample of nine. Then a control, a baseline check, two windows. Creative gets the same evidentiary standard as a bid change, which is rarer than it should be.
Send the current assets and the performance behind them. I’ll tell you what the creative is failing to answer, and whether that’s a production problem or a volume problem. It’s usually volume.
