The Retargeting Fallacy: A Case Study in Incrementality vs. Attribution

The Core Problem

Standard attribution models (specifically Last-Click) incentivize marketers to target high-intent users. This creates a feedback loop where budget is allocated to "harvesting" demand that would have converted organically.

High ROAS is often not a signal of efficiency, but a signal of selection bias.

Context & Hypothesis

I audited a luxury footwear retailer ($1,200 AOV) experiencing a revenue plateau despite reporting consistent high returns on retargeting spend.

• Market Variable: High brand penetration in the primary geo (Connecticut).
• Hypothesis: The retargeting campaigns were cannibalizing organic conversions rather than driving incremental lift. The customer decision cycle (2–4 weeks) was being interrupted by ads, not influenced by them.

Experimental Design

To isolate causal lift, we implemented a Randomized Controlled Trial (RCT) via Meta’s User ID segmentation.

• Treatment: 90% (Standard Retargeting)
• Control (Holdout): 10% (Suppressed from Retargeting)
• Constraint: A 50/50 split was statistically optimal but rejected by the client due to perceived revenue risk. The 90/10 split was the maximum viable "political" compromise.
• Duration: 14 days (Capturing 80% of the historical abandoned-cart-to-purchase window).

The Signal (Results) After the two-week window, the conversion rate delta between the groups was negligible.

• Group A (Ads): 5.2% Conversion Rate
• Group B (No Ads): 5.1% Conversion Rate
• Incremental Lift: +0.1%

The Economics of "Fake" ROAS While the platform dashboard showed a high ROAS based on attributed sales, the incremental math proved negative leverage.

• Ad Spend: $10,000
• Incremental Revenue Generated: ~$800 (derived from the 0.1% lift)
• Net Result: -$9,200 loss in 14 days.

The "proven" channel was effectively a tax on the brand's most loyal customers. Extrapolating this inefficiency over the client's 7-year strategy suggests approximately $280,000 in budget waste—capital spent to acquire customers who were already acquired.

Conclusion

There is a fundamental difference between Touch (attribution) and Lift (incrementality).

• Touch answers: "Did the ad exist before the sale?"
• Lift answers: "Would the sale have happened without the ad?"

If you cannot verify the latter, you are likely paying for your own organic success. The fix is not to optimize the creative, but to shift capital to prospecting—moving from harvesting convinced users to acquiring new ones