How We Achieved 6.2x ROAS Using Google Performance Max — Step-by-Step Campaign Breakdown

When our client — a D2C fashion brand based in Mumbai — came to us, they were running Google Smart Shopping campaigns with a 1.8x ROAS and burning through ₹15 lakh per month with little to show for it. Six months later, they were at 6.2x ROAS with the same budget. Here's exactly what we did.

The Problem With Their Original Setup

Before diagnosing what we'd fix, we ran a full account audit. The issues were textbook but costly:

• No audience signals: They were running broad Smart Shopping with zero first-party audience data fed in
• Poor feed quality: Product titles were generic ("Blue Dress" instead of "Women's Navy Blue Maxi Dress Summer Casual")
• No campaign separation: All products in one campaign — hero products competing for budget with slow-movers
• Zero creative testing: One set of images, one set of copy, running unchanged for 6 months
• Wrong bidding strategy: Maximise Clicks (should be Target ROAS)

Phase 1: Foundation (Weeks 1–3)

Product Feed Transformation

Everything in Performance Max starts with your product feed. We spent the first two weeks fixing theirs:

• Title formula: Brand + Gender + Product Type + Key Feature + Material + Colour + Size
• Descriptions: Added 500+ character descriptions with style guidance, occasion suggestions, and care instructions
• Custom labels: Tagged products by margin tier (High/Mid/Low), stock level, and season
• Image quality: Ensured all primary images had white backgrounds at 1:1 ratio; added lifestyle secondary images

Campaign Architecture

We split one bloated campaign into four focused ones:

1. Hero Products PMax: Top 20 SKUs by revenue — 60% of budget, Target ROAS 600%
2. New Arrivals PMax: Products under 30 days old — 20% of budget, Maximise Conversion Value
3. Clearance PMax: Slow movers on sale — 10% of budget, Target ROAS 300%
4. Brand Search Campaign: Separate search campaign for brand terms — 10% of budget

Phase 2: Audience Signals (Week 3–4)

Performance Max's "black box" reputation comes from marketers not feeding it audience signals. We uploaded:

• Customer email list (4,200 past purchasers) as a "Customer Match" audience
• Website visitors in the last 30 days from Google Analytics
• 30-day purchasers as a "high value" customer signal
• A custom intent audience built around fashion and competitor searches

Feeding these signals gave Google's machine learning a head start — instead of learning from scratch who to target, it started with a profile of your actual buyers.

Phase 3: Creative at Scale (Weeks 4–8)

This is where most brands underinvest. We built 5 asset groups for the Hero Products campaign, each with a distinct creative angle:

Asset Group 1: Lifestyle

Images of products being worn in real settings (cafes, streets, offices). Copy focused on occasions: "Perfect for date night" / "From office to evening"

Asset Group 2: Product Showcase

Clean product images with detailed descriptions. Copy focused on quality: "Premium cotton, not fast fashion" / "Made to last, not just to wear"

Asset Group 3: Social Proof

Images of customer reviews overlaid on product shots. Copy using real testimonials: "4.8 stars from 2,400 customers"

Asset Group 4: Sale/Value

Price-forward creative. Copy highlighting value: "Up to 40% off" / "Free delivery above ₹999"

Asset Group 5: Seasonal

Tied to current festivals or seasons. Copy aligned with occasion: "Diwali Collection" / "Summer Essentials"

Every 2 weeks, we pulled the performance breakdown by asset and replaced the bottom 20% with fresh creative.

Phase 4: Bidding Strategy Progression

The biggest mistake with Performance Max is switching to Target ROAS too early. We followed this progression:

1. Weeks 1–4: Maximise Conversion Value (no target) — let the algorithm learn
2. Weeks 5–8: Set Target ROAS at actual achieved ROAS minus 20% (conservative)
3. Weeks 9–12: Increase Target ROAS by 10–15% every 2 weeks as volume stayed consistent
4. Month 4+: Held at Target ROAS of 600% (6x), adjusting for seasonal demand

The key rule: never change your Target ROAS by more than 15% at a time. Larger changes throw the algorithm into re-learning mode and you lose 2–3 weeks of optimisation.

The Numbers: Month-by-Month Progress

5 Lessons for Your Performance Max Campaigns

1. Your feed is your foundation. Garbage in, garbage out. Spend 20% of your setup time on feed quality and you'll see 80% of the improvement.
2. Feed first-party audience signals from day one. Never launch a PMax campaign without your customer list and website visitor audiences loaded.
3. Separate campaigns by margin and intent. Putting high-margin hero products in the same campaign as clearance stock is a budget efficiency disaster.
4. Creative diversity matters. Google's algorithm needs different creative angles to find what resonates with different audience segments. Minimum 5 asset groups.
5. Be patient with bidding. The algorithm needs 4–6 weeks to learn properly. Resist the urge to tweak before it has enough data.

If your Google Ads are underperforming, the problem is almost certainly one of these five things. Fix them systematically and results follow.

Want us to audit your current Google Ads setup and identify exactly what's costing you ROAS? Get a free PPC audit here.