Skip to main content

Probabilistic Matching

Probabilistic matching relies on models to tie devices to individuals, households, or other groups. Associations between devices are not observed directly (unlike deterministic matching), but are inferred from mathematical models that evaluate the probability of one or more devices belonging to a given individual, household, or to another group (e.g., geography based).

Why Use Probabilistic Matching

The goal of probabilistic matching is to create device groupings with high numbers of linkages, enabling high cross-device reach use cases with relatively more false positives. This is dissimilar to what Google and Facebook offer, as it does not primarily rely on authenticated user traffic.

Since deterministic accuracy is traded for a high volume of linkages in probabilistic matching, these graphs are not suitable for use cases requiring people-based data. Instead, they’re useful when the goal is to target a large number of devices that have some chance of belonging to a single individual.

Advantages:

  • High number of cross-device relationships

  • Greater targeting reach

  • Lower complexity

Disadvantages:

  • Higher number of false positives

  • Does not provide people-based data

  • Less accurate measurement and targeting

  • Data lost over time: Device data are stored on device groups; device groups are created to store implicit device relationships and do not represent an individual, so an individual’s new and old device data may not be joined or stored over time

  • Does not accurately align with offline customer records

See our Cross-Device Methodologies document for more information on deterministic and probabilistic matching.