A 7-figure Shopify health brand cycled through three performance marketing agencies in 18 months. Each engagement followed the same arc: meaningful ROAS improvement in month 1 (2.4 → 3.1, then 2.6 → 3.4, then 2.9 → 3.6), plateau by month 2, decline by month 3, back to baseline by month 4. Same pattern, three different agencies. When we audited the account, the problem wasn't the agencies — it was that no engagement had a structured 90-day system. Improvement happened, then optimisation stopped because nobody had a clear model for what month 2 and month 3 should accomplish.
Why most performance marketing engagements stall at 60 days
Month 1 of any engagement produces results because it starts with an audit: identifying what's clearly broken and fixing it. Wasted spend cut. Creative refreshed. Landing page improved. These changes are high-leverage and produce real ROAS movement.
Month 2 is where the system is supposed to compound. Instead, most engagements shift to maintenance: reporting on what happened in month 1, running A/B tests without a clear hypothesis, adjusting bids. The improvement doesn't compound because the underlying system — the process for identifying constraints and fixing them — was never built. Month 1 was a one-time cleanup, not the beginning of a process.
Month 3: without a compounding system, ROAS reverts. The initial fixes have been absorbed. Creative that was refreshed in month 1 is now 60 days old and starting to fatigue. Nothing was built to replace it systematically. The landing page improved in month 1 but was never stress-tested at higher spend. The account looks like it did before month 1, except for a new set of changes that also need to be replaced.
The fix is not a better agency. It's a roadmap that specifies what the work is in each phase — so month 2 and month 3 have a defined purpose, not a continuation of month 1.
The 90-day arc: what should happen in each phase
Phase 1 (Days 1–30): Audit and stabilise.
The goal is not to improve ROAS — it's to understand exactly what's limiting it and to eliminate clear waste.
Specific deliverables:
- Full account audit: identify all ad sets with spend above $200 and zero purchases in 30 days. Pause them.
- Calculate break-even ROAS using Shopify P&L (COGS + fulfillment + platform fees + returns). This is the target, not "as high as possible."
- Establish baseline Shopify-attributed ROAS (UTM-filtered). This is the true number. Record it.
- Identify the primary constraint: is ROAS limited by CVR (landing page), creative (frequency, hook performance), or audience (size, exhaustion)?
- If the primary constraint is CVR: fix the landing page above-the-fold before week 2 ends. If creative: pause top-frequency assets, push budget to under-tested creative. If audience: address this in phase 2.
Phase 1 output: a primary constraint identified, a baseline Shopify ROAS documented, clear waste removed.
Phase 2 (Days 31–60): Fix the constraint and build the creative pipeline.
The goal is to address the constraint identified in phase 1 and to build the infrastructure that prevents month 3 regression.
Specific deliverables:
- If CVR was the constraint: landing page rebuilt and live by day 35. Test period days 35–50. Document new CVR baseline. Move to next constraint (usually creative or audience).
- Creative rotation: at least 3 new creative assets in testing by day 45. Each with a specific hypothesis. Weekly review of thumb-stop rate, hook-to-hold rate, landing page CVR per creative.
- Establish creative testing cadence: specific budget per test ($30–50/day), specific test duration (5 days), specific decision criteria (thumb-stop rate, hook-to-hold, landing page CVR). This is the process — not a one-time test.
- Audience expansion test: if audience size is the constraint, test Lookalike 1-5% against previous narrower structure.
Phase 2 output: constraint addressed, creative pipeline with at least 3 proven assets, testing cadence documented.
Phase 3 (Days 61–90): Scale what's working.
The goal is controlled budget increase — only after phase 2 has delivered above-break-even stability for 21+ consecutive days.
Specific deliverables:
- Confirm three criteria before any budget increase: CVR above 3.0% from paid traffic, at least 2 creatives with stable ROAS above break-even for 14+ days, account ROAS within 0.4 points for 21 days.
- First budget increase: 30–50%, not doubling. Hold for 2 weeks. Document ROAS stability.
- Second budget increase only if ROAS held at first increase. Same criteria check.
- Identify next constraint at new spend level (usually creative frequency or audience exhaustion — the ceiling moves when budget moves).
Phase 3 output: controlled scale with documented ROAS stability, and the next constraint identified for the next 90-day cycle.
How a health brand followed this arc and held 4.1 ROAS at month 4
Month 1: Audit found $8,200/month in wasted spend (legacy ad sets, audience overlap). Baseline Shopify ROAS: 2.4. Primary constraint: landing page CVR at 1.7%.
Month 2: Landing page rebuilt (above-the-fold rewrite, review count moved above fold, single CTA). CVR moved to 3.2% by day 45. Creative pipeline: 4 new assets tested, 2 cleared the 28% thumb-stop / 45% hook-to-hold thresholds. Creative testing cadence documented and handed to internal team.
Month 3: 3 criteria met for scale. Budget increased from $18,000 to $24,000/month. ROAS held at 4.1 through week 3 of month 3. Second increase to $29,000 in week 4. ROAS: 3.9 (within acceptable range). Creative frequency beginning to build — next constraint identified.
Month 4: 4.1 ROAS, stable. The system was running, not just the campaigns.
Map your current engagement against this framework
Wherever you are in your current engagement, ask three questions:
- Has the primary constraint been explicitly identified? Not "we're optimizing ROAS" — specifically: is CVR the bottleneck, or creative frequency, or audience exhaustion? If you don't know the answer, you're in month 1 mode regardless of how long the engagement has been running.
- Is there a documented creative testing cadence? A specific process — budget, duration, decision criteria — not just "we test creative." If not, month 3 regression is predictable.
- Are three scale criteria met before any budget increase? CVR above 3.0%, two proven creatives, 21-day ROAS stability. If a budget increase happened without all three, the scale is fragile.
If your current engagement doesn't have clear answers to all three, that's what to ask for before month 4 starts.