When this D2C apparel brand came to us, they were running Google and Meta independently — two agencies, no shared data, no unified creative strategy, and a blended ROAS of 1.8×. Revenue was growing, but ad spend was growing faster.
Ten months later, they were running a unified full-funnel programme across Google, Meta, and TikTok with a 4.6× blended ROAS and a cost-per-acquisition 62% lower than when we started.
Here's exactly how we did it.
The Starting Point
The brief was deceptively simple: improve paid media efficiency. But the real problem was structural. The previous setup had:
- Google Search and Shopping being managed separately from Meta, with no audience data shared between them
- Creative production happening on a six-week cycle, with no structured A/B testing — just creative "gut feel"
- Retargeting audiences based on 30-day website visitors, with no segmentation by engagement depth or intent signal
- No feed-based creative for Meta, meaning ads looked nothing like the actual product pages
- Attribution modelled on last-click, which was distorting which channels were actually driving revenue
The fix wasn't to spend more. It was to rebuild the architecture.
Month 1–2: Architecture Rebuild
Before touching a single ad, we rebuilt the measurement infrastructure:
- Server-side tracking: Implemented server-side events for both Meta CAPI and Google Ads conversion tracking. This alone recovered approximately 28% of previously unattributed conversions.
- Attribution modelling: Switched from last-click to data-driven attribution across Google, and implemented a custom attribution model in our reporting layer that gave partial credit across the full funnel.
- Audience segmentation: Built 12 distinct retargeting audiences based on engagement depth — from 25% video viewers to add-to-cart abandoners to previous purchasers by product category.
The measurement rebuild paid for itself in month one. Recovering 28% of lost attribution immediately improved the apparent ROAS of existing spend — without changing a single creative or bid strategy.
Month 3–4: Creative Testing Cadence
We introduced a structured creative testing framework. Every two weeks, we launched three new creative hypotheses — each tested with a minimum 5,000-impression budget before any scaling decision was made.
The creative variables we tested systematically:
- Hook format: text overlay vs. spoken hook in the first 2 seconds
- Social proof positioning: reviews at the start vs. at the end
- Product focus: single hero product vs. lifestyle context
- CTA language: "Shop now" vs. specific offer ("Free shipping today")
By month 4, we had identified three creative archetypes that consistently outperformed — and scaled spend into those archetypes, while keeping a rotation of new tests running.
Month 5–7: Full-Funnel Architecture
With clean measurement and winning creative in hand, we rebuilt the campaign architecture around funnel stage:
- Prospecting (top of funnel): Meta Advantage+ Shopping Campaigns for broad reach; Google Performance Max for intent capture; TikTok Top Feed for awareness and creative testing with a younger demographic.
- Consideration (mid-funnel): Meta retargeting of 3-second video viewers and 50%+ content engagers; Google Remarketing Lists for Search Ads targeting recent site visitors.
- Conversion (bottom of funnel): Dynamic Product Ads on Meta targeting cart abandoners and product viewers; Google Shopping Smart Campaigns with TROAS bidding anchored to our own blended target.
Month 8–10: Scaling and Optimisation
Once the architecture was performing consistently, we scaled spend — but methodically. We increased budgets by no more than 20% per week on any campaign to avoid disrupting the learning phase.
The TikTok channel, which started as a creative test environment, became a meaningful acquisition channel by month 9 — generating 22% of new customer volume at a CPM 40% lower than Meta.
What Drove the 4.6× Result
Breaking down where the ROAS improvement came from:
- Measurement recovery: ~0.8× improvement from server-side tracking and attribution rebuild
- Creative efficiency: ~1.2× improvement from systematic creative testing — winning creatives converted at 3.4× the rate of control creatives
- Audience architecture: ~0.9× improvement from segmented retargeting vs. broad 30-day pixel audiences
- Channel diversification: ~0.7× improvement from TikTok diversification reducing CPM inflation on Meta
No single change produced the result. It was the compound effect of all four working together.
Key Learnings
Three things we'd tell any D2C brand thinking about improving their paid media performance:
- Fix measurement before you fix anything else. You cannot optimise what you cannot accurately see.
- Treat creative like a product. Ship fast, test with discipline, scale what works. The intuition-driven creative cycle is the single biggest ROAS killer we see consistently.
- Full-funnel architecture is not a luxury. Top-of-funnel awareness without conversion infrastructure loses customers. Retargeting without prospecting dries up the pipeline.
If you want to talk through what a full-funnel rebuild looks like for your brand, book a 30-minute strategy call.