Probabilistic Matching

How Probabilistic Matching Works

Probabilistic matching is a device graph methodology that identifies likely relationships between devices without direct user authentication. Algorithms analyze patterns of shared signals — devices connecting from the same IP address, exhibiting similar browsing behavior, accessing the same apps, or frequenting the same physical locations — and assign probability scores to device relationships. For example, if a smartphone and a laptop consistently connect from the same home IP, visit the same news sites, and use the same time zone, the system assigns high probability they belong to the same person. Match rates for probabilistic models are higher than deterministic (sometimes 2–3x the reach), but accuracy varies — false match rates of 5–15% are common in consumer markets.

Why Probabilistic Matching Matters for B2B Marketing

Probabilistic matching has become more important in the post-iOS 14 era, where deterministic mobile identity signals are restricted. For B2B marketers, probabilistic matching enables cross-device reach extension beyond the subset of prospects who are logged in across devices. It's particularly useful for reaching prospects at the top of the funnel — where broad reach matters more than individual-level accuracy — and for extending reach in markets with limited deterministic data availability.

Probabilistic Matching: Best Practices & Strategic Application

When using probabilistic matching, validate performance through incrementality tests rather than view-through attribution, which can be inflated by false matches. Layer probabilistic cross-device reach with contextual and intent-based targeting to improve relevance. Always use frequency capping at the probabilistic cluster level (not individual device level) to prevent over-exposure when the matching model is uncertain. Regularly audit device graph vendors by testing known device pairs against their match rates.

Agency Perspective: Probabilistic Matching in Practice

MV3 uses probabilistic matching as a reach extension layer on top of deterministic foundations. In B2B campaigns, we set conservative frequency caps on probabilistic audiences (2–3 exposures/week vs. 4–5 for deterministic) to account for false match risk. This approach extends reach by an average of 40–60% beyond deterministic-only targeting while maintaining campaign efficiency within acceptable accuracy thresholds.