- Campaign architecture — not budget — is the primary driver of ROAS. Structure determines everything.
- Audience segmentation using first-party data and lookalikes consistently outperforms broad targeting.
- Creative fatigue is the #1 silent killer of Meta campaigns — systematic testing prevents it.
- Full-funnel campaigns (awareness → retargeting → retention) reliably outperform single-stage approaches by 2–4x.
Campaign Architecture: The Foundation
Most brands running Meta ads are fighting the algorithm instead of working with it. They set up a handful of ad sets, pick an audience, load a few creatives, and wonder why performance plateaus after week two. The issue isn't the creative. It isn't even the audience. It's the architecture.
The framework we use across all our D2C clients separates campaigns into three distinct layers — each with a clear objective, a defined budget ratio, and a separate measurement window. This isn't just about organisation. The separation tells Meta's algorithm exactly what you want from each campaign, allowing it to optimise independently for each goal rather than averaging across conflicting objectives.
The three-layer structure
- Prospecting (60% of budget). Cold audiences only — interest-based, lookalikes of your top 10% buyers, and broad with strong creative. Optimise for purchase with a 7-day click / 1-day view window.
- Retargeting (30% of budget). Website visitors (30 days), video viewers (75%+), and add-to-cart abandoners. Optimise for purchase or initiate checkout. Exclude recent buyers.
- Retention (10% of budget). Past purchasers 90–365 days. Push complementary products or upsells. This is your highest-ROAS layer and it's chronically underfunded by most brands.
Audience Segmentation That Actually Works
The death of third-party cookies and iOS privacy changes have made broad audience targeting more viable, not less — but only when paired with exceptional creative. The brands still obsessing over granular interest targeting are leaving performance on the table.
Our approach to prospecting audiences in 2026:
- Lookalikes of your highest-LTV customers — not all purchasers, specifically your top-spending 10%. The difference in quality is significant.
- First-party data custom audiences — CRM lists, email subscribers, app users. Upload as a seed for lookalikes and exclude from prospecting.
- Broad + Advantage+ audiences — let Meta find the buyers. Works best with a proven creative set (5+ winning ads) and a strong Pixel event history.
"The brands that refuse to let Meta's algorithm do its job — by over-constraining audiences — consistently underperform those that give it room to optimise."
— Mystiq Media Performance TeamCreative Testing: The Engine of Consistent Performance
Creative fatigue is the silent killer of profitable Meta campaigns. An ad that generates a 10x ROAS in week one can drag down to 2x by week four — not because the audience changed, but because the same people have seen it too many times. Most brands don't have a system to catch this before it costs them.
Our creative testing protocol:
- Launch a minimum of 6 new creative concepts per month per account — 3 static image variants and 3 video variants.
- Test within the prospecting campaign with a capped budget per ad set to allow fair comparison.
- Declare a winner at 500 impressions minimum and a
frequency > 1.5. Kill the losers fast. - Winners graduate to the main prospecting campaigns. Losers are analysed for learnings and iterated on.
- Archive and refresh winning creatives every 6–8 weeks even if performance hasn't dropped — proactive rotation beats reactive rescue.
Measurement: What ROAS Number Are You Actually Tracking?
The most common mistake we see when auditing underperforming accounts is measuring the wrong thing. Platform-reported ROAS (Meta's native number) is almost always inflated — sometimes by 40–60% — because of attribution overlap between channels and the way Meta credits view-through conversions.
We track three numbers simultaneously for every client:
- Platform ROAS — what Meta reports. Useful for relative comparison within the platform, useless for cross-channel assessment.
- MER (Marketing Efficiency Ratio) — total revenue divided by total ad spend across all channels. The most honest number.
- New Customer ROAS — revenue from first-time buyers only, divided by prospecting spend. This is the number that determines if your acquisition economics are sustainable.
The 8.4x ROAS figure we cite comes from MER — total tracked revenue against total Meta spend, across a 90-day rolling window. It is not the platform number. This is an important distinction: it means the actual economic return is real, not a reporting artefact.
If you want us to audit your current campaign structure and show you where your ROAS is leaking, get in touch — we do this as part of every onboarding engagement.