From Blind Spending to Margin-Positive Growth.

The Client

A French home & décor e-commerce brand with a broad product catalogue, selling across France and Western Europe. They were spending €150,000 per month on Google Ads alone — over €200,000 per month across Google, Meta, Microsoft, and TikTok combined — primarily through Performance Max campaigns, legacy search campaigns, and Shopping — and had been growing their ad budget year-on-year without a clear picture of what was actually working.

The brand came to us with a straightforward ask: explain why ROAS feels wrong, and fix it.

The Challenge

When we ran the initial audit, we found three compounding problems:

• Conversion tracking was measuring revenue, not margin. The primary conversion action feeding Smart Bidding was tracking gross merchandise value (GMV) — not the brand’s actual gross margin. Smart Bidding was optimising for the wrong number, inflating apparent ROAS while masking genuine profitability.
• Performance Max was running unconstrained. 96 asset groups across the main PMax campaign had no target ROAS set. Two campaigns were operating below the 3.0x margin ROAS breakeven point, burning money with no automated brake.
• Browse-trigger products were invisible to Smart Bidding. A significant portion of the catalogue drove high browsing traffic and contributed to sessions that ended in purchase — but because the purchase was attributed to a different product, Smart Bidding was suppressing these “browse trigger” products. We suspected this was hiding substantial value.

The Solution

We approached the rebuild in three phases:

Phase 1: Conversion tracking rebuild. We identified the correct gross margin conversion action and re-mapped it as the sole Smart Bidding signal. We documented the break-even point (3.0x margin ROAS) and aligned all optimisation logic around it.

Phase 2: Full account restructure. We rebuilt the campaign architecture across PMax, Shopping, Search, and Demand Gen around a single 3.0x margin ROAS break-even target. High-margin products were prioritised at every level. All 96 PMax asset groups received explicit tROAS constraints aligned to product category margin rates. We fixed 27 asset groups with broken final URLs pointing to discontinued sale pages, and paused the two confirmed below-breakeven campaigns.

Phase 3: Browse trigger pipeline. We built a custom attribution pipeline joining Google Ads product-level data with GA4 BigQuery session sequences. This let us identify products that consistently appeared early in browsing sessions that ended in purchase — and calculate their true session-level contribution value.

• GA4 BigQuery export analysis to reconstruct CPC session paths
• Product-level ROAS recalculation: direct vs. session-assisted ROAS
• Automated monthly pipeline to refresh browse trigger scores
• tROAS recommendation engine with confidence thresholds

The Results

Within 60 days of implementation, the data told a materially different story:

Beyond the headline numbers, the brand now has a measurement architecture they can trust. Every euro of ad spend maps to a margin outcome. The browse trigger pipeline runs automatically on the 1st of each month, surfacing new bid recommendations without manual analysis.

What’s Next

We are currently extending the engagement to cover Meta Ads and Microsoft Ads, applying the same margin-first measurement framework across channels. A cross-channel attribution model is in development to bring the full paid media picture into a single Looker Studio dashboard, updated daily via BigQuery.