Meta Ads Account Architecture for DTC | Apex Growth Corp
Paid Media · Account Architecture

Meta Ads Account Architecture for DTC: The System Behind 35.69× ROAS

S
By Sayed Sultani, Founder
·June 2, 2026·5 min read

Key takeaways

Most DTC Meta accounts aren't a system. They're a pile of campaigns — one launched for the spring sale, three more from a freelancer two years ago, a "test" that quietly became 40% of spend, and a retargeting campaign that's just re-selling people who already bought. Spend goes in, some revenue comes out, and nobody can say why.

When we inherited a stalled apparel account stuck at roughly 2× blended ROAS, that's exactly what we found. Ninety days later it was running at 35.69× blended ROAS — $19.6K in spend returned $700K+, across 24,379 purchases at ~$0.80 cost-per-purchase. The creative mattered. The offer mattered. But the thing that made all of it legible — the thing that let us scale without the account falling apart — was the architecture. Here's how we build it.

Why the default structure fails

The instinct most brands (and a lot of agencies) follow is to optimize at the ad-set level: spin up more ad sets, set a ROAS target on each, pause the losers, duplicate the winners. It feels rigorous. It's actually the problem.

Fragmenting budget across dozens of tiny ad sets does three destructive things at once:

The architecture: three layers, organized by intent

We rebuild every account around a single principle — structure follows intent, not ad format. Three layers, each with a different job and a different success metric.

Layer 1 — Prospecting (cold)

This is the engine. Its job is to find net-new buyers profitably, and it's where the majority of budget lives. We run it broad and consolidated: a small number of high-budget campaigns using Advantage+ Shopping and broad targeting, letting Meta's signal do the audience-finding instead of slicing into a dozen hand-built interest stacks. Fewer, better-fed ad sets. The success metric here isn't last-click ROAS — it's new-customer cost and blended account ROAS.

Layer 2 — Mid-funnel (engaged)

People who engaged but didn't convert — video viewers, site visitors, add-to-carts that never checked out. This layer is small, surgical, and offer-driven. It exists to close the people prospecting already warmed up, not to inflate a vanity number.

Layer 3 — Retention & retargeting (warm)

Existing customers and high-intent visitors. Tightly capped budget. We treat its reported ROAS with deep suspicion — it's the layer most likely to claim credit it didn't earn — and we judge the whole account on blended performance, not this campaign's inflated line item.

"Judge the account on blended ROAS. The moment you optimize each campaign to its own attributed number, you've started lying to yourself."

Consolidation beats fragmentation

Once intent is mapped, the next move is ruthless consolidation. We collapse redundant ad sets, kill the long tail of micro-budgets, and concentrate spend where the system has enough signal to actually learn. A clean account that a stranger could read in sixty seconds will almost always outperform a "sophisticated" one with sixty ad sets nobody fully understands.

Creative velocity is the real lever

Architecture makes spend legible; creative is what makes it scale. Once the account is fed cleanly, the constraint becomes hook fatigue — the same winning ad decays as frequency climbs. The only durable answer is volume.

On the apparel account we tested 40+ creative concepts in 60 days — UGC, motion, statics, founder-voice, problem-agitate, social-proof — killed the losers fast on a clear threshold, and poured budget into the handful that broke out. You are not looking for one perfect ad. You're building a pipeline that reliably produces the next winner before the current one dies.

Attribution: fix the signal before you trust the number

None of this works if your data is lying. Post-iOS, browser-side pixel tracking under-reports and mis-attributes constantly. Before we scaled a single dollar, we rebuilt measurement:

Better signal feeds the algorithm better data, which improves delivery, which improves results — a compounding loop most accounts never even start.

The 90-day rollout

Architecture isn't a one-day rebuild you flip on. Here's the sequence we ran:

The Result · E-Comm Apparel
35.69×
Blended ROAS
$700K+
Conv. value
24,379
Purchases · ~$0.80 CPP

Will every account hit 35.69×? No — that brand had a strong product, healthy margins, and an offer that converted. Architecture doesn't manufacture demand that isn't there. But it's the difference between a brand that has demand and can't scale it profitably, and one that compounds. The structure is what lets everything else — creative, offer, budget — actually pay back the next dollar.

Want this run on your account?

We'll tear down your current Meta structure live, map where intent and budget are leaking, and show you exactly what we'd rebuild — in a free 60-minute audit.