Identity Resolution vs. Entity Resolution

Entity resolution is the capability. Identity resolution is the person-level slice of it. I use the two terms that way because the graph is built that way: one capability that works across people, households and companies, and one part of it that works on a single person.

The two terms get used interchangeably. That hides the one question that decides whether a result is right: what kind of thing is being resolved?

What entity resolution does

Entity resolution decides that separate records describe the same real-world thing. In our graph the thing is a person, a household or a company.

In data engineering the same phrase also describes cleaning duplicates out of a master record. It is the same idea at a smaller scale: deciding that two records are one thing. What changes here is the span. One graph resolves people, households and companies together, from signals that arrive in different places.

A company example: Cisco

Cisco has an office at location A, and traffic from it arrives on IP address X. Cisco has another office at location B, on IP address Y. Two locations and two IP addresses look like two visitors to a system that reads only the address.

Entity resolution ties them together as one company. A visit from either address returns the same answer: Cisco. That is a company-level answer. It tells you which company, and it does not tell you who.

A person example

A person is a bundle of identifiers that never arrive together: a work email, a personal email, a home address, the logins on an Apple TV and an iPad. None of them says it belongs to the same human as the others. Resolve them and they form one identity.

Identity resolution is the person-level slice of that work. A browser cookie resolves to a hashed email, and the hashed email resolves to a named profile. The step from cookie to hashed email is a deterministic match, which means the join either exists or it does not. Nothing is inferred.

The rule that keeps the two apart

An IP address resolves to a company or to a household. It never resolves to a person. This is where the two terms matter most, because a company is not a person, a household is not a person, and an answer at one level cannot stand in for another.

Households sit between the two. An IP address can resolve to a household at address level, and a household connects to the people who live there and to their mobile advertising IDs. Those are separate steps: the IP address itself still never resolves to a person. If a vendor tells you an IP address is a person, ask how the match was made.

What to ask when you buy

Two systems that both say they resolve identity can be answering different questions. These show which:

• Which entities do you resolve: people, households, companies, or only one of them?
• At what level does an IP address resolve?
• Which identifiers join to make one person, and is each join deterministic?
• Does each answer say whether it is a person, a household or a company?

The mechanics underneath are in what identity resolution is. For how a match rate is measured, read the honest numbers. Once a person is resolved, intent is read against that person: what they research, and when, against their own baseline.