Identity Resolution: Stitching User Profiles Together

WHY IT EXISTS: Users interact with a business across many channels: a website, a mobile app, a support chat, or a physical store. Each channel often generates data with its own identifier, like a browser cookie or device ID. Without a way to connect these disparate pieces, a business sees multiple fragmented users instead of one single customer, leading to flawed analytics and a poor, disjointed experience.

THE MENTAL MODEL: Think of identity resolution as a detective building a case file on a single person. Each piece of data, like an email address, phone number, or device ID, is a clue. The goal is to link these clues from different sources (your app, your CRM, third-party tools) to a single, unified identity graph. This master profile connects all known identifiers and interaction histories to one person.

HOW IT WORKS: The core task is called record linkage, which happens in two main ways. First, deterministic matching links records using exact matches on stable, unique identifiers like a hashed email address or a government ID. This is highly accurate but misses links when data is inconsistent. Second, probabilistic matching uses algorithms to score the likelihood that two records refer to the same person, even with slight variations like misspelled names or different addresses. This finds more matches but increases the risk of false positives. Most modern systems use a hybrid approach, starting with deterministic links and then carefully applying probabilistic models to expand the identity graph.

WHEN TO USE IT: Use identity resolution when you need a single view of the customer. This is the foundation for Customer Data Platforms (CDPs), advanced personalization across web and mobile, and accurate calculation of user-level metrics like lifetime value. It is also essential for complying with privacy regulations like GDPR, which require you to find and manage all data associated with a specific user upon request.

WHEN NOT TO USE IT: Avoid complex identity resolution when the cost of a false positive merge is catastrophic. In sensitive domains like healthcare or finance, incorrectly merging two people's records can have severe legal and personal consequences. It's also unnecessary overhead if your systems are already built around a single, consistent user ID that is enforced everywhere.

ONE CANONICAL EXAMPLE: A user visits your website, creating a record with cookie_abc. Days later, they sign up for your newsletter on their phone, creating a separate record with email_xyz. Initially, these appear to be two different users. Later, the user logs into the website using email_xyz. A deterministic rule can now link cookie_abc and email_xyz to the same person. The system merges their profiles, connecting their previous anonymous browsing history to their new user account, enabling a personalized experience.