Optimizing personalization in digital ads: modeling the trade-off between targeting breadth and creative depth across the marketing funnel

This research examines how personalization width (audience targeting, marketing funnel stage) and personalization depth (creative complexity, retargeting logic) jointly influence e-commerce advertising performance. Using a year-long dataset from Meta Ads and Google Analytics, 24 campaigns were systematically varied in a quasi-experimental field setting across funnel stages, audience types (custom, lookalike), retargeting strategies (standard, dynamic), and creative formats (low vs. high personalization). Predictive modeling using Random Forest and linear regression identifies click-through rate and audience size as the most consistent predictors of Return on Ad Spend (ROAS). The results indicate that low-complexity creatives can outperform highly personalized ads in top-funnel campaigns, while dynamic retargeting yields superior ROAS in middle- and bottom-funnel stages, particularly for custom audiences. An exploratory clustering analysis highlights three distinct risk–return campaign portfolios, illustrating how different personalization configurations balance performance efficiency and volatility across the funnel. The findings advance personalization and advertising optimization research by demonstrating the conditional and nonlinear effects of targeting and creative strategies, and by framing campaign design as a portfolio allocation problem in data-driven digital marketing.