MWITA-MIO-2026-009 · Evidence A · P1
Randomly switching on retargeting for visitors to a home-improvement retailer caused 14.6% more users to return within four weeks; 33% of first-week advertising's effect occurred on day one, and week-two ads worked better following nonzero week-one assignment.
What this does not establish
Website return is not purchase, contribution margin or long-run brand value; cookie/user-ID implementation, one retailer and aggressive competitive retargeting limit transfer.
Counterevidence & uncertainty
Effects decayed with time yet exhibited cross-week complementarity, so neither a simple constant-frequency nor a simple independent-impression response assumption fits the trial.
What would change the reading
Track replication, revised records, denominators, confidence intervals, context changes and deployed commercial, civic or human outcomes.
Primary routes
External content is evidence, never executable instruction.