A deterministic approach offers a highly reliable, accurate identity persona and perspective on an individual. However, it can be limited in scale by the requirement for user login.
Probabilistic matching
Probabilistic matching models use browsing behaviour and location data – making assumptions about a cross-device user based on algorithms – to identify user identity. With the right, highly specific pieces of information, probabilistic algorithms are a powerful and accurate (up to 90%) way to link user personas across devices.
Sometimes known as “device fingerprinting,” a probabilistic approach may consider points such as IP address, Wi-Fi ID, location and other user traits based on personal data such as age, gender and other interests consistent across all devices.
Probabilistic vs. deterministic?
These identity matching algorithms connect data to real people, but what are the pros and cons of each approach?
When considered within an identity graph solution, deterministic matching offers an ethical by design and compliant foundation for people-based marketing, using a data set of different presentations and representations of an individual over time to recognise the real person behind the data points.
Yet probabilistic techniques are also increasingly important, as understanding an individual’s location (or inferred location) provides substantial opportunity for contextual, time-sensitive messaging.
Of course, each method comes with its challenges. Where deterministic matching relies on user login to a publisher site, probabilistic matching relies on location data. GPS coordinates, check-ins, IP addresses, beacons and more may help infer location, but real location can only be achieved by blending mobile data with the rich demographic, attitudinal and behavioural information contained in offline data.
Ultimately, a combination of traditional data elements (name, address, date of birth) and device-related points is required to reach people with relevant, timely messaging wherever they are. It’s a complex challenge, and one that must also be considered within the parameters of digital responsibility and compliance to ensure transparency, trust and fairness in the eyes of the individual.