Sponsored brands and display ads on Amazon are often judged by surface metrics like click-through rate and cost-per-click, but the real financial impact shows up in the margin waterfall. In this episode, Lucas and Luna unpack how attribution windows and cross-channel cannibalization create a blind spot that can silently drain your total ROAS. They break down a scenario where a 20 percent ROAS on sponsored display hides a net loss of 8 percent after marginal cost of goods sold and overlapping organic sales. Using a specific hypothetical with a supplement brand spending $80,000 per month, they show how to build a blended ROAS model that accounts for attribution delay, view-through conversions, and cannibalization. Lucas explains why the 14-day attribution window is too short for consideration-stage products, and Luna challenges whether float-on-brand is ever worth the spend. The episode ends with a concrete framework: run a 30-day attribution holdout, calculate the cannibalization rate, and shift budget to bottom-of-funnel search if the blended ROAS drops below 2.5x. No fluff, just the math.