Instead of relying on cookies, they lean on contextual cues like page content, keywords, and real-time bidding context. They also use probabilistic models that group users into cohorts based on shared behaviors, device attributes, and geo/time patterns. First-party data from logged-in users and server-side conversion events helps fill gaps left by blocked scripts. Aggregated measurement, incrementality testing, and marketing mix models refine which placements actually drive signups. Over time, these systems learn patterns without needing full tracking, adjusting bids and creatives dynamically.