Dubai-based DTC apparel brand
+4.7× ROAS on Meta in 60 days
A premium apparel brand with $8M annualized revenue was plateauing at a 2.1× Meta ROAS despite creative output that looked great. We found a creative-fatigue + audience-overlap problem, rebuilt the ad ops as an AI-automated weekly cadence, and pushed ROAS to 4.7× without touching the brand direction.
The situation
The brand had great product, great photography, a solid influencer flywheel — and flat ROAS. Every new creative drop performed well for 8–10 days, then died. The team was spending 60% of its time editing variants and re-uploading, not on new ideas.
What we found
Two compounding problems:
-
Audience overlap. The Meta account had 14 active saved audiences. Overlap analysis showed 42% of impressions were landing on users already in 3+ of those audiences. In practice, the brand was paying to show three ads to the same woman in the same week. Frequency was high, CTR was decaying accordingly.
-
Creative fatigue, not creative shortage. Plenty of raw photography. The bottleneck was the variant-generation + copywriting + approval cycle — it took 4 weeks from "new hero asset" to "ad live in market." By the time the variant shipped, the audience had already seen it 6 times on the original.
What we shipped
Week 1: a weekly n8n workflow pulls audience-overlap data from the Meta Marketing API, writes a consolidation proposal, and posts it to Slack for approval. Approve → auto-restructure the campaign.
Week 2–3: an AI agent reads each hero asset + brand guidelines and generates 12 variant briefs (hook, copy, crop suggestion, CTA variants). Team approves what survives. A second agent produces the actual variants. Approval flow lives in Slack — no ad-ops dashboard to log into.
Week 4: integrated with a shared Notion library so the brand + creative teams have full visibility into what's shipping when.
Weeks 5–8: iteration. Refined the agent prompts against performance data. Cut the review burden on the team by another 50%.
The numbers now
ROAS is holding at 4.7× after 3 months. The creative team is spending its time on new product photography, not on ad variant drudgery. The founder stopped getting 11 PM "why is CAC up this week" texts from the CFO.
What's next for this client
Phase 2 is expanding the same automation to Google Shopping + a new Pinterest push. Same pipeline, new channels.
“”
Your pipeline, next
Could be your number we’re writing about next quarter.
More work
US-based mental-health platform
62% of support tickets resolved by AI chatbot
A teletherapy platform was drowning in Tier-1 support tickets — scheduling, password resets, insurance questions. We designed an AI chatbot on their stack (HIPAA-safe), integrated into their portal + SMS channel, and shifted 62% of weekly ticket volume off the human team in 90 days. Zero escalations, human handoff when needed.
UK-based B2B SaaS (DevOps tooling)
+412% organic traffic in 9 months
A mid-market DevOps platform had plateaued on paid acquisition and couldn't justify more spend to the board. We audited their technical SEO, rebuilt the content strategy around comparison + problem-first queries, shipped 38 deep articles in 9 months, and grew non-branded organic sessions 5.1×. CAC from organic now sits at 1/6th of their paid equivalent.
Morocco-based B2B SaaS
+3.2× qualified pipeline in 90 days
A post-Series-A scheduling SaaS with €2.4M ARR had flat outbound reply rates and a 28-day sales cycle that kept slipping. We rebuilt their inbound + outbound as a single automation, cut the cycle to 14 days, and tripled qualified pipeline without increasing headcount.