How I work through a growth brief, end to end
This is a work sample rather than a client engagement. It was a structured brief with five parts, each mapping to something the job actually involves. I’ve kept the method and taken out the brand names and absolute spend figures, because those accounts weren’t mine to publish.
It’s a fair picture of how I think when I’m handed an account and a question.
Part one: make raw exports say something
The brief. Take raw Meta performance exports for a 7 day and a 30 day window. Pull out cost, cost per install and ROAS at D3 and D7. Identify the top and bottom performers. Summarise the insight.
What I did. Cleaned and transposed both files into one sheet with a consistent structure: campaign, cost, CPI, ROAS D3, ROAS D7. Then I categorised every campaign against D3 ROAS into four buckets: scale, scale up cautiously, cut or pause, and ignore for now because the sample is too small to say anything.
The bucket that matters most is the fourth one. Most reporting quietly treats low-volume campaigns as if their numbers mean something. They don’t, and acting on them is how budget gets wasted on noise.
I also added a ReadMe tab documenting every transformation and the logic behind each threshold. If someone else picks up the sheet in three months, they can tell what I did and disagree with it. That’s the point.
Part two: a weekly performance review a client can act on
The brief. Review a live Facebook account for a DTC nutrition brand and write the weekly client update. The objective was minimising cost per add-payment-info.
What I did. Structured it the way I’d want to receive it:
- Overview. Where the account stands in one short paragraph.
- Insights and recommendations. What the data says, and the specific next action for each point. Not observations, decisions.
- Top and bottom ad sets. With the call attached: scale, hold, or cut.
- Performance breakdown. Campaign, then ad set, then creative level.
The gap between the best performing ad set and the account average was large enough that the recommendation wrote itself: concentrate budget, cut the tail, and rebuild the test structure so new creative gets a fair read.
I won’t publish the client’s spend or CPA numbers here. They aren’t mine to share.
Part three: creative briefs grounded in motivation
The brief. Review current creative trends, analyse two app products, and write four new creative concepts rooted in real user motivation.
What I did. For each product I wrote down the audience and, more importantly, what they actually want, before touching a concept.
For a halal dating app, the audience was Muslim professionals in their late twenties and thirties looking for marriage in line with their values. The motivations: an authentic connection, shared values, and profiles they can trust. That produced two concepts, one built on the long-term partnership promise, one a split-screen from first chat to first meeting.
For a casual card game in LATAM, the motivations were daily progression, social bragging rights, and nostalgia for a card game people grew up with. That gave a comeback-from-losing concept and a daily streak concept built around the habit loop and leaderboard pride.
Concepts written from motivation transfer between accounts. Concepts written from “what’s trending” don’t.
Part four: the difficult client conversation
The brief. A client is giving vague feedback, has rejected several rounds of creative, and the designer is taking the brunt of it. Handle it.
What I did. Two things are true at once here. The client needs reassurance that the work is moving. And the project is getting more expensive every day it drags, which is a real problem that reassurance alone doesn’t solve.
So: one more round with tightly scoped feedback, and a written note to the client afterwards confirming exactly what will change and what won’t. Then, separately and privately, a conversation with the designer to acknowledge it’s been rough, make clear I’m protecting their bandwidth, and offer to move them off the account if they want out.
After that, a tighter internal QA step so work gets reviewed through the client’s eyes before it’s shared, and a note on the account flagging it as high-sensitivity for future capacity planning.
Vague feedback is usually a symptom. Someone hasn’t articulated what success looks like, often because nobody asked them to.
Part five: competitor research at scale
The brief. A competitor is running tens of thousands of ads. Extract the creative insight without reading all of them.
What I did. Laid out three routes depending on budget and engineering access.
The quickest is a creative intelligence tool that already indexes the ad libraries. You get hook frequency, visual patterns, and which variants are being tested against each other, without building anything.
The middle route is scraping ad copy into a CSV, then clustering the hooks by emotional and functional category with an LLM to surface the recurring structures. For video, transcribe the top twenty and extract the narrative patterns the same way.
The most durable route, if there’s engineering time, is building it in house against the Ads Library API. Fetch the creatives, transcribe them, run them through an analysis step, and get back content buckets and fresh concepts on a schedule rather than as a one-off.
The competitor I looked at ran five clear hook categories: curiosity, social proof, a time-based challenge, direct comparison against alternatives, and a persona-led narrative. Every ad carried subtitles, high-contrast thumbnails and a CTA overlay, with the same script tested repeatedly against different visuals and voices.
That last detail is the actual lesson. They weren’t producing thousands of different ideas. They were producing a handful of ideas and testing the variables around them properly.
The thread running through all five
Every part comes back to the same instinct: work out what the numbers actually support before deciding anything, then write the decision down so someone else can run it.
If that’s how you’d like your account handled, book a call.
Brand names and absolute spend figures removed. The method, structure and reasoning are unchanged.