fb-pixel
Back to Case Studies
Case Study 03Premium Skincare D2C — Agentic Campaign Management8 min read

AI tools everywhere. ROAS stuck at 1.2x for six months straight.

A premium skincare D2C already running copy-gen, creative-gen, and a media buyer. On paper: modern stack. In reality: ROAS stuck at 1.0–1.3x for six months, POAS negative, and a founder who couldn’t pinpoint where the margin was going.

Industry
Premium Skincare D2C
Channels
Meta + Google
Time to result
Month 1
ROAS uplift
+115% in 30 days

The story in 10 seconds

A premium skincare D2C brand had every modern AI tool — copy-gen, creative-gen, a media buyer — yet ROAS sat at 1.2× for six straight months and POAS was negative: it lost money on every ad rupee. The tools generated content; nobody ran the campaign logic. We put four agents to work — killing boosted-post spend, pruning zero-intent search, refreshing creatives on real fatigue, and routing budget by margin. In the first 30 days, ROAS more than doubled to 2.6× and POAS turned positive (+14%) for the first time in the brand’s history.

01 — The Setup

Every AI tool running. Nothing connecting them into a system.

The brand had invested in the right tools — a copy-gen tool for ad creatives, a creative-gen tool for visuals, and an experienced media buyer. Every individual piece looked modern.

But ROAS had flatlined between 1.0x and 1.3x for six consecutive months. POAS was negative — the brand was losing money on every ad rupee spent. The AI tools generated content. Nobody was running the business logic that should have connected them.

Content generation is not campaign management. One produces assets. The other ensures those assets actually make the business money.

1.2x
Average ROAS over 6 consecutive months
40%
Meta budget on boosted posts optimized for likes
−8%
POAS — losing money on every ad rupee
Monthly
Creative refresh cadence, regardless of actual fatigue

02 — What Was Broken

Four silent drains. None of them obvious without the right layer.

0140% of Meta spend was on boosted Instagram posts — optimized for likes and reach, not add-to-carts or purchases. Budget burning on vanity metrics.
02Google Ads bleeding on completely zero-intent searches: “DIY face pack,” “home remedies for pimples” — audiences who would never buy a premium skincare product.
03Creatives refreshed on a fixed monthly calendar — not when they actually fatigued. Weeks of budget wasted on stale ads while the team waited for the next scheduled refresh.
04The highest-revenue SKUs were also the lowest-margin SKUs. ROAS looked fine on these. POAS was deeply negative. Nobody was routing budget based on margin.

ROAS & POAS trend — 6 months stuck, then Flable

Budget allocation before vs. after Flable

03 — What Flable AI Did

Four agents running together as one system.

C

Creative Fatigue Agent

Monitored performance signals continuously. Swapped in new creatives the moment engagement dropped — not on a monthly calendar, but in real time when fatigue was detected.

K

Campaign Restructure Agent

Killed all boosted-post spend entirely. Rebuilt Meta campaigns under proper Sales objective with full pixel event tracking, conversion APIs, and SKU-level audience targeting.

S

Search Hygiene Agent

Auto-pruned negative keywords on Google every week, continuously — no human review cycle, no monthly ritual. Zero-intent search terms removed automatically.

P

POAS Routing Agent

Shifted budget toward SKUs that were genuinely profitable after margin. Revenue increased. Margin increased. Both simultaneously, for the first time.

04 — Outcome by Month 1

From margin-negative to profitable — in the first 30 days.

ROAS
↑ Month 1
Stuck at 1.0–1.3x
2.6x

+115% improvement in 30 days. Six months of flatline reversed in one month.

POAS
↑ First ever
−8% — losing money
+14%

From bleeding margin on every ad rupee to genuinely profitable. First positive POAS in brand history.

Google Wasted Spend
↓ Pruned
Irrelevant zero-intent keywords
−32%

Wasted keyword spend pruned; Google CTR up 1.8x on remaining high-intent terms.

Creative Refresh Cycle
↓ Automated
Monthly calendar
9days avg

Cycle compressed from 30 days to 9 — automated and triggered by actual fatigue signals.

ROAS, POAS, and Google CTR index improving together after Flable

Bottom line

AI tools generate content. Flable AI runs the business logic. There’s a meaningful difference.

Frequently Asked Questions

We already use AI tools for copy and creative. Why isn't that enough?

Those tools generate assets; they don't run campaign management. More creative doesn't fix budget sitting on boosted posts, zero-intent Google keywords, or margin-blind spend. Flable runs the business logic — what to spend, where, and when to refresh — against profit, not output volume.

What's the difference between ROAS and POAS — and why did POAS flip the picture?

ROAS is revenue ÷ ad spend; it can look healthy while you lose money, because here the highest-revenue SKUs were the lowest-margin ones. POAS (Profit on Ad Spend) accounts for margin and cost, so routing budget by POAS turned a −8% into +14% — profitable for the first time.

How can creatives refresh faster than monthly without more work?

The creative-fatigue agent watches performance signals continuously and swaps creatives the moment engagement drops, instead of waiting for a calendar date. The cycle compressed from 30 days to about 9 — automatically.

Will this replace my media buyer or my tools?

No. It sits on top of them and runs the connective logic — campaign structure, search hygiene, fatigue detection, and POAS routing — so your team and tools do their best work against the right metric.

What do you need to get started?

Read-only access to your Meta and Google accounts. Flable analyses where spend is leaking and shows your real ROAS and POAS on your own data, typically within 48 hours.

Still refreshing creatives on a monthly calendar?

Free 30-minute audit. We’ll analyse your Meta + Google account and show you exactly where the spend is leaking — on your own live account data.

Book a free audit →

Setup in 48 hours · No credit card required · Backed by Microsoft & NVIDIA