The Conversion Goal That Counted Add-to-Carts as Sales.

Two numbers and a board meeting. That is what the operator brought to the audit.

Claimed ROAS, twelve-week trailing, was 3.8x. The Google Ads dashboard said the operator was making three dollars and eighty cents in revenue for every dollar spent. The agency summary said the same thing. Both numbers were consistent with each other.

Shopify total revenue, same window, told a different story. Total revenue had grown four percent year over year. The Google Ads account had grown spend twenty-eight percent. If claimed ROAS were correct, Shopify revenue should have moved more than four percent. The math did not close.

Actual purchase ROAS, when reconstructed from Shopify-side purchase data only, was 1.4x. Less than half of what the dashboard said. The dashboard had been right about the conversion count and wrong about what the conversions were.

The board meeting was in three weeks. The CFO asked the marketing lead to reconcile the two checks. The marketing lead asked Stan Consulting to assess the account before the meeting.

The audit was scoped at this point. Seventy-two-hour written verdict. The brief was one sentence: tell us why the dashboard says we are profitable and the bank account says we are not.

Add-to-cart was a primary conversion goal. That sentence is the verdict. Smart Bidding was optimizing against a combined signal that counted intent as outcome.

The defect was a single line in the conversion-goal grouping. The default Shopify-Google integration via GTM had pushed three events to Google Ads as conversions: view_item, add_to_cart, and purchase. The agency had imported all three. View_item had been demoted to secondary. Add_to_cart and purchase had been left in the primary grouping. The bidding algorithms used the combined primary signal as the optimization target.

Five things were true at the same time. Each one was a consequence of the goal-grouping defect. None of them were independent.

One. The combined primary conversion goal counted add-to-carts and purchases as a single signal. A purchase carries an add-to-cart event with it; the same user produced two "conversions" for the algorithm. The claimed conversion count was inflated by the add-to-cart-only completions where the user added to cart and abandoned. The algorithm could not tell the two apart because they shared the same primary-goal status.

Two. Smart Bidding learned that traffic with high add-to-cart propensity was high-value. It optimized toward that traffic. The audience segments the algorithm preferred were users who add things to carts and abandon · deal-shoppers, comparison-shoppers, list-builders. The campaign got cheaper "conversions" and lower-rate purchases simultaneously, which is the structural fingerprint of this defect.

Three. Claimed revenue was double-counted at the conversion-value level. The default GTM tag had attached a value parameter to the add_to_cart event equal to the cart's monetary value at the moment of add. When the same user purchased, the purchase event also fired with its own value. The combined primary goal summed both. The dashboard assess $3.80 in revenue per dollar spent because it was counting the same dollar twice for converters and once for cart-abandoners.

Four. Performance Max compounded the issue more than Search did. PMax leans heavily on whatever signal it is told to optimize against because the campaign type has fewer manual controls. Without a clean primary conversion, PMax optimized against the inflated combined goal more aggressively than the Search campaigns did, and PMax was the campaign with the most spend. The bigger the budget on PMax, the bigger the misallocation.

Five. The gap between claimed ROAS and actual purchase ROAS widened the longer the campaign ran. After twelve weeks the algorithm had fully converged on the wrong target. The campaign was systematically buying clicks from users who would add to cart and not purchase, because that is what the bid signal said was a conversion. More learning made the misalignment more efficient.

The audit's written verdict named the install order. Order matters. Re-pointing Smart Bidding before fixing the conversion goal would have stripped the campaign's learning history and re-trained against the same corrupted target.

The audit drove into the Conversion Audit engagement format and from there into a thirty-day install. The conversion-goal cleanup below is what was installed.

Each link below points at a related Atlas page that handles a piece of the case file in more depth. Reference pages give the definition. Position pages give the firm's defended doctrine. The hub gives the map.