Ammeil Ali — Creative Lead Portfolio

How I've improved creative production and performance

Three systems, not three habits

Most creative teams get faster by working harder. These are the three pieces I build so output scales without quality sliding, and so every brief can point at the thing that justified it.

01 · RESEARCH

Creative DNA

Before any brief, the brand gets a research workbook: customer reviews, Reddit and Amazon mining, competitor ad-library teardowns, and sentiment analysis, resolved into personas, awareness stages, angles, and 100 scored hooks.

Every brief cites a source, not a hunch.

02 · PRODUCTION

The pipeline

Briefs flow through Notion to designers, editors, and creators on a fixed cadence, with locked naming, versioning, ratio, and delivery conventions. AI handles volume on statics and concepting; strategy and QC stay human.

3x output at the same headcount.

03 · QA & LEARNINGS

The feedback loop

QA checklists for briefs, statics, and video catch problems before the client sees them. Biweekly partner reviews hold quality. A weekly learnings note turns last week's reads into next week's test roadmap.

Quality holds flat as volume climbs.

Managing the people who make it

How I hold a partner to a bar

Agencies, freelancers and in-house teams all fail the same way: the brief was vague, so the work came back defensible. I manage that with written non-negotiables and explicit rejection criteria, agreed before anyone starts.

EXTERNAL AGENCIES

Avasant

Managed external agency partners at a global management-consulting firm serving Fortune 500 clients, while leading an internal team of four. Enterprise procurement, enterprise standards, enterprise consequences for a missed deadline.

Agency and vendor management, two years.

TEAMS & FREELANCERS

GrowthQuest

Direct in-house designers, editors and creators plus freelance partners and outside production vendors across a portfolio of brands, setting priorities, running the review cadence, and owning whether the output ships.

Multiple partners, multiple brands, concurrently.

THE MECHANISM

Written, not verbal

Every partner works from a brief carrying the deliverable spec, the non-negotiables, and the criteria that get a batch sent back. Biweekly review on quality and pace. A QA checklist clears it before a client ever sees it.

Nothing is rejected on taste alone.

Real brief · production partner, Sept 2026

The non-negotiables from a live video brief

Verbatim from a brief I issued to a production partner for six UGC video ads. This is the section that decides whether a batch is accepted or sent back.

The product must match the real unitCondition on our supplied photos, never a text description. No restyling, recolouring or reshaping. "This is the one rule that gets a batch rejected. The client sells a $4,200 product and buyers purchase what they see."

No text inside generated framesGenerated type warps and misspells, so every headline, review card and CTA is added in the edit from supplied copy. A past render produced "WARBIOR WILLPOWER". A misspelled wordmark cannot ship.

One consistent face per adSame person from first shot to last. Most tools re-roll the face every clip, so lock a character reference first and drive every shot from it.

It has to look like a phone, not a productionHandheld, available light, ordinary rooms with clutter. No studio lighting, no grade, no flare. If it looks produced it reads as an ad and the format stops working.

Why it is written this way

Each rule names the failure it prevents and cites the batch that failed. A partner can self-check against it before delivery, which is the difference between a revision round and a rejection, and it means feedback is never a matter of my opinion against theirs.

Case study 01

Building the creative engine behind a 331% climb

Consumer health · recovery hardware

Warrior Willpower, hyperbaric chambers, cold plunge & saunas

High-ticket DTC, 14-month engagement. My role: creative lead, working alongside a separate media buyer and the client's founder. I owned research, briefs, production, concept QA, and the biweekly review cadence.

$972K+Revenue generated

331%Monthly revenue growth

5.85xAverage ROAS

9.88xPeak ROAS

The account

Fourteen months on a high-ticket recovery-hardware brand selling hyperbaric chambers, cold plunges, and saunas at roughly a $3.3K AOV. Monthly revenue climbed from $43K to $188K, a 331% increase, while monthly spend scaled from $12K to $25K+ and ROI stayed positive the whole way. Holding efficiency while spend doubles is a creative problem long before it is a bidding one, which is why the creative engine is the story here.

The creative engine, and the month it produced

I built the full AI-assisted creative system end to end: market research resolved into a Creative DNA, distinct angles mapped across the funnel, then brand-true statics generated at volume and QA'd for launch. 30+ statics across 4 product lines and 9 angles, with no studio and no photoshoot.

That launched set returned 9.2x blended ROAS and $188K revenue on ~$20K spend, the account's best month of the entire engagement.
The single best creative, a hyperbaric-chamber concept, drove $100K+ in revenue at 10.6x ROAS on its own.
Top ad of the set peaked at 13.99x. 57 purchases at ~$3.3K AOV.

Then it stalled, which is the part worth reading

Later in the engagement the account softened: ROAS slid from 6.47x to 5.74x month over month and CPA climbed from $399 to $490, on a month that still turned over $129,298. Growth had come from volume, and the library had drifted back toward product education. What follows is how I read it and what I briefed next.

What I did

Pulled every top performer of the period and read them against spend, ROAS, and CPA rather than against opinion, then wrote up five repeatable directions the winners shared. The month's best single concept had done $38,173 at 7.79x alone, so the question was what it carried that the rest did not.
The directions: calendar-moment urgency over evergreen offers, outcome-led headlines instead of spec lists, the discount as a hard number on a visible badge, real-life emotional framing over product-only shots, and a direct second-person voice.
Briefed the next month's pack against those five, manufacturing a calendar moment where the real calendar gave us none.
Held two hard production rules: exact product representation, QA'd side by side against the client's own unit photos, and distinct design concepts only, never copy variations of one layout.
Delivered every concept at 1:1 and 9:16 to spec, so nothing got cropped at launch.

The pack those five directions produced

August had no holiday to sell against, so the calendar moment was manufactured. Every unit below carries a visible discount badge, leads on outcome rather than spec, and speaks in second person. Concepts are split by product, never all-products-together.

Ice bath ad reading Summer Ends. You Don't — **Calendar urgency.** Manufactured deadline, second-person voice.

Couple by an ice bath, Recover Smarter This Fall — **Outcome-led.** Relationship imagery, not a product-only shot.

Group cold plunge ad, Join the 6AM Cold Club — **People and trust.** Community, faces, star rating.

Hyperbaric chamber ad citing Good Morning America — **Authority.** Press mention plus a 4.9 rating.

Annotated hyperbaric chamber ad, Your Mornings, Upgraded — **Annotated benefit.** Outcome mapped onto the product in use.

Chiller ad reading Vacation's Over — **Product-specific.** Chiller gets its own concept, not a bundle.

What the data actually said

The two highest-ROAS ads of the period both used relationship and lifestyle imagery with visible trust cues, not product-only renders. Product education, with no offer or urgency, did not compete in feed at any spend level. That single read redirected how the following month's concepts were weighted.

Why the discipline mattered

Near-identical visuals accelerate fatigue and give the buyer nothing to read, since one layout in four tones tests tone, not concept. Enforcing distinct concepts is what kept a small weekly volume producing clean signal.

Video, UGC & scripts

Briefed, scripted, shot, cut

Statics are half the job. The other half is scripts, hooks, UGC and edits, judged on hook rate and hold rate before anything downstream matters, because an ad nobody watches past three seconds has no CTR worth reading.

Scripting · consumer health

Six UGC scripts for a $4,200 hyperbaric chamber

Hook / Body / CTA compressed to UGC length, ~130 spoken words at 150wpm. One distinct reason to buy per script, no shared body with swapped hooks. Persona and awareness stage named on every one, with the angle traced back to a verbatim customer review in the brand's research workbook.

How a script is laid out

Two columns, so the editor is never guessing. Left is what is on screen, right is what is said. Overlays and runtime markers are specified inline rather than left to interpretation.

Handheld selfie, kitchen, no setup. Talent already mid-thought."My mom's had arthritis in her hands for about fifteen years."

Cut to chamber in her spare room, zipped, lamp on."She said this is the first thing that's actually helped her. No side effects."

Small self-correction, hand up."And look, it wasn't overnight. Took her about three weeks before she said anything to me."

Text overlay: 60-DAY TRIAL · FREE RETURN SHIPPING"It's $4,200. I'm not going to pretend that's nothing. But there's a sixty day trial."

Runtime map on every script: hook 0–5s, body 5–45s, CTA 45–60s. Delivered 9:16 and 1:1, both required, because 4:5 gets cropped at the edges on launch.

Finished video · multiple formats

Produced work

Street interview, owner-voice direct response, and UGC talking head. Different formats because they answer different objections, not because variety looks good in a deck.

Street interview approach shot — **Street interview, open.** Real passerby, unscripted reaction.

Street interview subject holding a flat-fee card — **Street interview, payoff.** The offer card lands in frame.

Owner voiceover ad filmed on a job site — **Owner voice.** Five angles scripted for the founder, shot on site.

UGC talking head filmed at home — **UGC talking head.** Phone-shot, captions burned in.

Watch one

The street-interview cut in full, 60 seconds. Briefed, shot list set, and cut to land the offer in the final beat.

Case study 02

Concept discipline, shown

One insight, four genuinely different executions, one locked offer bar. This is what I mean when I say distinct concepts rather than copy variations, and it is the standard every brief I write is held to.

Retail services · seasonal push

"Open on Holidays"

Insight from review and competitor research: on public holidays every competitor closes, and phones break anyway. One line of customer frustration became the whole campaign.

Shuttered storefronts at night with one lit repair shop — **Competitive contrast.** Everywhere else closed, one light on.

Technician repairing a phone at a workbench — **Proof of work.** A real technician, mid-repair.

Cracked phone screen showing a calendar — **Problem-first.** Damage meets the calendar.

Miniature beach scene on a cracked phone screen — **Cultural moment.** Local idiom, playful diorama.

How the set is built

Four separate visual strategies: competitive contrast, proof of work, problem-first, and cultural moment. A buyer can tell which idea won, not just which tone.
One message per creative. The headline carries the whole argument, the offer bar carries the detail, and nothing competes in the middle.
Offer bar, hours, branch, and handle locked identically across all four, so the only variable being tested is the concept.
Every unit shipped at the launch ratios with the logo, safe zones, and legibility checked against the QA list before delivery.

The production system

AI does volume. A person keeps strategy and QC.

What it is

The engine behind case study 01, now running across the whole portfolio: an in-house AI creative pipeline of 35+ custom production skills that takes a researched brief to finished statics and UGC concepts the same day, QA'd to brand-designer standard. Strategy, art direction, and the final pass stay with a person, which is the only reason the output is usable.

What it changed

A new angle goes live the same day it is identified, instead of waiting on the next production sprint.
40+ assets a month across statics, video, and UGC, at 3x the previous output with no added headcount.
No studio, no photoshoot, and no drop in brand standard, which is what makes 9 angles across 4 product lines affordable in the first place.

The rule that keeps it honest

Anything generated is QA'd against the real product before it ships. On high-ticket categories especially, people buy what they actually see, so a render that flatters the product into something the brand does not sell is a defect, not a creative choice.

Reporting & learnings

The monthly creative read

A standing analysis that rolls every spending ad up by concept and scores the concepts, not the ads. This is what turns a month of output into next month's test roadmap, and it is the document the team and the client actually argue over.

Creative performance analysis · OpusClip

Concept-level reporting on a venture-backed SaaS account

Every ad name encodes its creative concept, so a 30-day window of 21 spending ads rolls up into five concepts ranked on cost per signup, CTR and 3-second view rate. Long-tail dormant ads excluded so the read is not diluted.

What the read produced

The winner was starved. Trending / Product Demo led on every metric, best cost per signup, best CTR, best 3-second view rate, and held only 3% of spend. Named as the single biggest scale opportunity in the account.
The loser looked healthy. Claymation carried a strong view rate and a cost per signup 57% above account average. The verdict written into the report: the hook works, the message doesn't. That is a rewrite instruction, not a pause instruction.
Separating hook performance from conversion performance is the whole point. One says stop scrolling, the other says buy, and a concept can win the first and lose the second.

Cadence

Monthly on creative, with country, audience-segment and placement breakdowns running alongside it, plus separate TikTok audits where the account runs there. Each one ends in a ranked list of what to scale, what to rewrite, and what to cut.

Absolute spend and conversion volumes withheld. Relative figures and verdicts are as written in the original.

Range

Two more registers, same process

Direct response is one voice. The brief system is the same whichever voice the brand needs, because the research decides the register, not my preference.

Premium apparel · editorial

Featherie, women's golf apparel

Founder-led brand in a category dominated by men's cuts. The research said the buyer was tired of being an afterthought, so the work is restrained and editorial, with the offer badge deliberately absent.

Navy golf polo editorial ad, Built for her game — **Positioning.** "Built for her game. Not borrowed from his."

Raspberry polo and skort editorial ad — **Use case.** Course to clubhouse, no outfit change.

Sage green golf set ad, Performance you can feel — **Product truth.** UPF 50+, real pockets, four-way stretch.

Four-panel collage of golf apparel — **Range.** Collage for bestsellers and catalogue intent.

Woman and girl in matching golf apparel — **Audience widening.** Women and girls in one frame.

Founder in branded sweatshirt on the course — **Founder story.** Started at 14, tired of boys' cuts.

Lead generation · proof-led

Van Media, lead gen for moving companies

An audience that had been burned by lead vendors before. Claims were worthless, so every concept is built on a screenshot of the actual outcome, with the guarantee stated in the headline stack.

Ad showing an inbox full of new leads — **The inbox.** Volume shown, not claimed.

Ad showing a booked calendar — **The calendar.** Booked jobs, not raw leads.

Ad showing a payouts dashboard — **The bank.** Revenue as the closing proof.

Testimonial-style ad — **The peer.** Someone like them, saying it first.

Why the ladder works

Inbox, calendar, bank. Each asset answers the objection the previous one leaves standing, so the set functions as a sequence rather than four interchangeable images, and the buyer can read which stage of doubt is actually costing conversions.