Specialty retail, high-ticket ecommerce engagement.

Situation

A specialty retailer selling high-ticket ecommerce products at an average order value of approximately $1,000 had been running Google Ads with an incumbent agency for an extended period. On the surface the account looked functional: spend was consistent, conversions were registering, and ROAS was sitting at 2.47. But revenue growth had plateaued despite increasing spend, and the client had no clear explanation for why.

The account was handed off with no audit documentation and no baseline on non-brand performance.

Diagnosis

A full account audit surfaced five structural problems:

1. Consent Mode V2 misconfiguration

The consent management platform (CMP) had been implemented incorrectly, causing approximately 30% of event data to go unmodeled. Smart Bidding was optimizing against an incomplete signal from day one.

2. Duplicate conversion actions

Lead form submissions and purchase events were being counted across overlapping tags, inflating conversion volume and distorting CPA figures across every campaign.

3. GA4 audiences not connected to Google Ads

Remarketing lists, customer match segments, and high-intent behavioral audiences built in Google Analytics had never been piped into Google Ads. The account was bidding blind on all warm traffic.

4. Weak product feed

Titles were generic, key attributes were missing, and there was no segmentation by product type, margin tier, or sale price. PMax was running broad across the entire catalog with no structural guidance.

5. Brand cannibalization via PMax

Branded search terms had not been excluded from Performance Max campaigns. The account was paying a 20% cost premium to capture traffic that would have converted through brand search — inflating overall ROAS while suppressing true acquisition performance.

Intervention

Work proceeded in a deliberate sequence. Tracking before structure, structure before bids.

Phase 1 — Tracking stabilization (Days 1–30)

Consent Mode V2 reconfigured at the CMP level. Duplicate conversion actions audited, deduplicated, and consolidated into a single source of truth. GA4 audience segments connected to Google Ads and applied across relevant campaigns.

Phase 2 — Structural rebuild (Days 31–60)

Product feed rebuilt from the Shopify source — titles rewritten with primary attributes leading, segmentation added by product category and margin tier. PMax campaigns restructured around product segments rather than a single catch-all. Brand exclusions applied to PMax to eliminate cannibalization. Dedicated brand Search campaign isolated with controlled budget.

Phase 3 — Learning and stabilization (Days 61–120)

New campaign structure passed through Smart Bidding learning period. Bids held stable deliberately — no target adjustments until clean conversion data had accumulated. ROAS reached breakeven at 2.8x by day 120.

Phase 4 — Optimization (Days 121–180)

Bid strategy calibrated against verified conversion data for the first time. Non-brand ROAS isolated and tracked as the primary performance metric. Spend scaled incrementally as efficiency gains confirmed.

Result

By the end of the engagement period:

• ROAS improved from 2.47 to 3.57 — a 44% efficiency gain
• Conversion value increased by $31,920
• Conversions up 306 units
• Spend grew 45% with revenue growth outpacing spend

Why it worked

Every prior optimization attempt had been built on corrupted inputs. Consent Mode misconfiguration meant Smart Bidding had never seen the full conversion picture. Duplicate tags meant nCAC (new customer CPA in-platform) figures were fictional. GA4 audiences sitting disconnected meant warm traffic was being treated as cold. None of those problems were visible in the agency’s reporting.

Fixing the measurement layer before touching campaign structure or bids meant that when the rebuild happened, it happened on a foundation that could actually hold the result.