What Is Identity Resolution? Deterministic and Probabilistic Matching, Explained
Last reviewed: 2026-08-07
Identity resolution is the process of linking the fragmented identifiers a person leaves across devices, browsers, and systems, such as an email hash, a cookie, a device ID, and a loyalty account, into a single persistent profile of that person or household. It is the connective layer that lets a brand recognize the same customer across a website visit, an app session, and an offline purchase.
What it does
Identity resolution takes identifiers that arrive disconnected, an email address from a purchase, a cookie from an ad click, a device ID from an app open, and builds a graph that links the ones belonging to the same person. That graph then answers a specific question on demand: given this identifier, what else do we know is the same person. Systems downstream, a CDP, a DSP, a measurement platform, query that graph rather than rebuilding the matching logic themselves.
What it does not do
Identity resolution does not collect the underlying data itself; it links data that other systems already hold. It does not decide what to do with a matched profile, whether to suppress an ad, personalize a message, or exclude someone from a campaign; that decision sits in whatever system consumes the resolved identity. And it does not guarantee certainty. Every graph carries some rate of false matches and some rate of missed matches, and the honest evaluation question for any identity provider is what that rate looks like on data resembling a brand's own, not whether matching happens.
Deterministic versus probabilistic matching
Deterministic matching links identifiers only when there is an exact, verified point of connection, such as the same hashed email appearing in two data sets or a login tying a device to an account. It produces fewer matches but each one carries higher confidence.
Probabilistic matching links identifiers using statistical similarity, such as shared IP address, device characteristics, and behavioral patterns, when no exact key connects them. It produces more matches, extending the graph to logged-out sessions and devices that never shared a deterministic key, at the cost of some false-positive rate that grows as the graph reaches further from verified connections.
Most identity providers blend both: a deterministic core built from verified identifiers, extended by probabilistic linking where deterministic data runs out. The mix a provider uses, and how it discloses confidence at each layer, is a real point of difference between providers, not a settled industry standard.
Where it fits in a stack
Identity resolution sits underneath the systems that need to recognize a person, rather than being the system a marketer works in directly. A CDP relies on it to unify a profile across sources that do not share a common key. A DSP relies on it to hold targeting and frequency capping together across cookieless and identified environments. A measurement platform relies on it to connect an exposure to an outcome across devices. Buyers most often meet identity resolution as a component inside one of those tools rather than as a product bought on its own, though dedicated identity providers exist and often supply the graph that several of those downstream tools query.
Among the more visible dedicated providers: LiveRamp operates an identity graph and data collaboration tools used to connect and activate matched identifiers across partners. Acxiom provides identity and data linkage services spanning offline and online sources.
Related reading
- How to Evaluate a CDP: What Separates the Platforms, where identity resolution quality is one of the core evaluation criteria.
- How to Evaluate a DSP: What Separates Programmatic Platforms, a system that depends on resolved identity for targeting and frequency capping.
Frequently asked questions
Is identity resolution the same as a device graph? A device graph is one output of identity resolution, specifically the set of devices linked to the same person. Identity resolution is the broader process, which can also link non-device identifiers such as email hashes and loyalty account IDs into the same profile.
Does identity resolution work without cookies? Deterministic identity resolution built on emails, logins, or loyalty IDs does not depend on cookies and continues to work as third-party cookies phase out further. Probabilistic matching that relies partly on cookie-based signals loses some of its input as cookies decline, which is part of why deterministic identifiers have become more central to identity strategy.
How accurate is identity resolution? Accuracy varies by provider and by how far a match sits from a deterministic identifier. A match built on a verified email is close to certain; a match built on shared device characteristics across a large probabilistic graph carries a measurable false-positive rate. Ask any identity provider for their match-rate and confidence methodology rather than a single accuracy number.
Is identity resolution the same thing as a CDP? No. A CDP unifies and activates customer data, and identity resolution is one capability it relies on to do that unification across sources that do not share a common key. Some CDPs build identity resolution natively; others license it from a dedicated identity provider.